Corps de l’article


At least since Bentham, there has been a well-known and influential criticism of exclusionary rules in evidence law.[1] The criticism goes something like this: people have a natural way of coming to knowledge, which does not use formal rules to exclude some pieces of information from consideration. Instead, people freely consider all relevant information and give each piece the weight it deserves based on its reliability and degree of relevance. Information may still be excluded from consideration, but it is excluded after a judgment is made as to its particular characteristics and value. Evidence law artificially excludes much relevant and valuable information on the basis of rules that are partially doing this assessment for the fact finder. These rules may be designed to avoid the prejudices of juries, but that could effectively be accomplished by a more careful screening and training process.[2] Furthermore, such rules certainly should not apply to judges themselves.

The influence of this criticism cannot be overstated.[3] Even writers who argue against Bentham tend to do so by taking issue with its practicality,[4] saying that while an ideal juror might be trainable to give all relevant information its proper weight, actual jurors are too deficient and the resources of time and money are too limited to perform the necessary training. On the other hand, judges themselves see the Benthamite critique as a reason for them to ignore the rules of evidence when they are sitting at bench trials,[5] considering themselves to be closer to the ideal fact finder and believing themselves more able to assign all relevant information its proper weight.[6] This is reinforced by the wide latitude afforded judges in implementing the rules, in many cases by the rules themselves.[7]

In this paper I argue that the push toward judicial discretion and “universal admissibility”[8] (also called “free proof”[9]) has costs of which reformers were unaware. Moving toward a universal admissibility standard and increasing judicial discretion sacrifices the promise that legal conclusions will be reached on a uniform standard of knowledge reproducible across cases. This in turn jeopardizes the promise of justice, especially in criminal trials where defendants are not being tried according to uniform justificatory standards.

While these are the ultimate practical worries, the bulk of the paper will be devoted to developing the epistemological argument that the inclusion of certain kinds of information, the prime example being hearsay, can lower the standards for what counts as knowledge, essentially making it too easy to justify attributions of knowledge. In that, it is a philosophical argument with a practical conclusion once the impact of the argument on evidentiary practices is understood. The claim is that there are objective but variable standards of justification for knowledge and that one factor in this variability is the kind of information presented to the fact finder. Hence, given our desire for uniformity in the standards for knowledge across cases, evidentiary exclusions like hearsay are useful tools for limiting the variability in justification. We want to maintain high standards for what is to count as knowledge in the courtroom, and evidentiary exclusions can help us do that.

A few caveats are in order before we begin.[10] There has been plenty of ink spilled over discoveries in the psychology of how people treat evidence and the application of these discoveries to evidence law.[11] For the most part, these studies and their implications are not directly relevant to the points advanced in this paper. Since this paper is about the implications of epistemological theory for evidence law, the point it makes does not hang upon the psychological ability or inability of fact finders to treat evidence in a certain way. Rather, the point is that there are objective epistemic standards for what is to count as knowledge, which evidence law should be seeking to replicate in determinations of admissibility.

This is not to say that the reliability of information is irrelevant to admissibility, but that the epistemic standards at play ought also to be an important factor in fashioning and deploying the rules of admissibility. While empirical psychological research may point away from the utility of certain exclusions, I argue that there are still important philosophical reasons for maintaining those exclusions.[12]

Another caveat is that the argument presented is mainly aimed against the trend toward finding more hearsay to be admissible. However, the philosophical points raised here could also be used to criticize the relaxations of certain other exclusions. In the United States, for example, Federal Rule 903 has abolished the need for a subscribing witness to authenticate documentary evidence unless the jurisdiction otherwise requires it (some U.S. states, for example, still require attestation for wills to be admissible[13]). This allows documentary evidence to speak for itself in much the same way that jurors are likely to encounter such documents outside the courtroom. Similarly, Federal Rule 1003 allows for reliable duplicates of documents to be admitted instead of originals unless there is a “genuine question” raised about the authenticity of the originals. Federal Rule 1004 through 1007 provide for other ways to present the content of documents when originals are not available or otherwise not necessary.

We can say that these exceptions and those discussed below with regard to the admissibility of hearsay derogate from the best evidence rule.[14] Since one could see the best evidence rule itself as partially providing for the special epistemic context of the courtroom, such exceptions and limitations will tend to blur the epistemic line between the courtroom context and the mundane epistemic contexts we usually find ourselves in outside the court.

I. Bentham’s Critique

Bentham classified evidence law as “adjective” law, grouping it with procedure and distinguishing it from “substantive” law.[15] As a part of procedure, its primary object was to allow decisions to be reached that conform to substantive law with a minimum of the “inconveniences ... of delays, vexations, and expense.”[16] However, Bentham saw an essential tension between substantive law and adjective law in that the latter can be used to reach a conclusion that would contradict what the substantive law seems to promise.[17]

Bentham himself notes that the habitual reception of reliable information by others in daily discourse induces a general “disposition to believe” the testimony of others.[18] Of course there are sufficient instances of deception for there also to be a disposition to doubt, but this is the exception and depends upon a specific cue that causes doubt.[19] In general, the normal machinations of society depend heavily upon the assumption that most testimony of others is truthful.[20]

That said, Bentham does classify hearsay as “makeshift” evidence, indicating that it is an inferior form of evidence for judicial purposes as a result of the inability of a party to cross-examine the declarant.[21] Given this inferiority, Bentham advises a basic rule of not admitting hearsay evidence.[22] He also, however, articulates a sweeping exception that swamps the rule.[23] He would admit the hearsay whenever examination of the declarant is “either physically or prudentially impracticable,” where the latter includes a “preponderant inconvenience in the shape of delay, vexation, and expense.”[24]

In this we can understand his considerations as stemming from a form of the best evidence rule.[25] Bentham’s main concern leading him to allow the bulk of hearsay to be admitted into evidence is the worry that information not otherwise attainable would be excluded from consideration, and the claim that this result is worse than the risks of unreliability associated with the evidence itself.[26] That is, the harm of a “misdecision” based on incomplete information is greater than the harm of a “misdecision” based on imperfect information, the dangers of which can be cured by retrial when made manifest.[27]

II. Impact of Bentham’s Critique

Nineteenth-century reformers largely ignored Bentham’s arguments on hearsay even while accepting many of his other reforms.[28] At the turn of the twentieth century, however, the tide began to shift.[29] In the United States, this culminated in the American Law Institute’s Model Code of Evidence proposed rule 503(a) in 1942: “Evidence of a hearsay declaration is admissible if the judge finds that the declarant ... is unavailable as a witness.” Their rationale for this rule is that it explicitly

treats the jurors as normal human beings, capable of evaluating relevant material in a court-room as well as in the ordinary affairs of life [where] they hear, consider and evaluate hearsay.[30]

As we will see, it is the very ubiquity of knowledge based on hearsay outside the courtroom that renders it problematic for knowledge attributions in the higher stakes context of the courtroom. Nevertheless, these first salvos were not, at first, received positively by courts and legislatures.[31] Reformers then turned to professing a rejection of the Benthamite principle of universal admissibility where the declarant was unavailable, even as they whittled away at the hearsay exclusion by using a series of expanding exceptions.[32]

The focus has for many years been upon those increasingly byzantine exceptions, leading many jurisdictions to adopt some form of catch-all or principled exception.[33] In the United Kingdom, hearsay is admissible in civil cases[34] and it is admissible in criminal trials when the declarant is unavailable or in fear of bodily harm or financial loss for giving testimony, or when the court deems it in the interests of justice to do so.[35] In Canada, where hearsay exclusions and their exceptions are based on common law principles and court decisions, the increasing complexity of implementing common law exceptions has led courts to embrace an “increasingly flexible” approach termed the “principled approach”.[36] This has led to a shift in the basic assumptions about hearsay, leading some to argue that the emphasis is now on inclusion unless there are “compelling reasons to exclude”.[37] While legally the presumption is still on exclusion, there is no question that the principled approach has led to a much broader admissibility of hearsay and represents a large step toward free proof.[38] As will become apparent, there is some irony in the notion that the principled approach allows courts greater adaptability in their sensitivity to contextual considerations surrounding the creation and admission of the out-of-court statement or document.

In Khan, the Supreme Court of Canada introduced what became known initially as the principled exception, allowing the out-of-court statement of a child where “the guarantees of necessity and reliability are met.”[39] This was understood to include non-infant declarants by the Court in Smith.[40] In that decision, the Court also articulated guidelines for the criteria of reliability and necessity: following Wigmore, reliability is understood as a “circumstantial guarantee of trustworthiness” in the facets of the context in which the statement was made.[41] Necessity is understood as the statement being needed to prove any fact at issue.[42] In quoting Wigmore’s understanding of the aspects of what makes an out-of-court statement necessary to include situations in which “[t]he necessity is not so great; perhaps hardly a necessity, only an expediency or convenience, can be predicated,”[43] the Court hearkens back to Bentham’s preference for inclusion where there would otherwise be a “preponderant inconvenience in the shape of delay, vexation [and] expense.”[44]

In R. v. B. (K.G.),[45] the Court extended the principled approach to situations in which the declarant is also available in court. More recently, in Khelawon, the Court elaborated on the reliability requirement noting that it is met when the circumstances show that there is no concern about whether the out-of-court statement is true,[46] or where the fact of it having been made out of court is not of concern since there are other means of testing its veracity than by cross-examination.[47] The Court also notably allowed “all relevant factors” to be considered in assessing threshold reliability, “including, in appropriate cases, the presence of supporting or contradictory evidence,”[48] thus overturning restrictions[49] on the use of extraneous evidence in determining admissibility stemming from Starr.[50]

What is clear from these developments is that courts and reformers, once they overcome concerns with the rights of the accused to confront witnesses, are primarily concerned with the reliability of the information obtained from the out-of-court statements.[51] As will become clear, however, this criterion is not enough to cover all of the considerations that need to be addressed when dealing with the justifications for beliefs of fact finders in the courtroom.[52]

III. Knowledge Attributions

Among its other tasks, the law of evidence should be concerned with ensuring that knowledge attributions made by fact finders to witnesses or to themselves are made on the basis of adequate and reproducible justifications.[53] That is, when fact finders hear a witness give testimony, they are called upon to make a decision about whether to believe the witness is telling the truth and whether she knows the information to which she is testifying, and then to incorporate that information into their own corpus of beliefs after entertaining beliefs in contradicting information. Hence, an analysis of evidentiary rules should not be solely concerned with the relevance and reliability of the evidence admitted, but also with the way that evidence suggests the justifications for a fact finder’s beliefs.[54]

Some may counter that the object of evidence given at trial and hence of evidence law is that the beliefs of fact finders have a high probability of being true.[55] On this view, probabilistic belief rather than knowledge is therefore the object of evidence law; the reliability of information presented is the only concern. While a complete reply to this objection would require the paper be devoted to the subject,[56] there are a number of problems with this view that can be listed here to motivate the claim that knowledge attributions ought to be understood as the main focus of evidence law.

First, it is not clear that the idea of probabilistic belief is actually at odds with a focus on knowledge attributions.[57] If knowledge is approximated with true beliefs that are sufficiently justified,[58] then probabilistic belief claims are merely claims that the proper justification is a high degree of probability. Second, note that probabilities themselves must be beliefs in order to be acted upon by the fact finder.[59] Under this view, “[t]o justify a finding, the fact-finder needs only to believe, and be justified in believing, in the relevant proposition of probability.”[60] Hence, if the object is a justified belief in probability, we are still in the realm of determining what justifies knowledge attributions about those propositions of probability. Notice as well that the goal of beliefs that have a high probability of truthfulness is not the same as the goal of admitting only reliable information. Anyone concerned with ensuring a high probability of true beliefs will need to investigate further how information presented factors into belief formation and whether reliability can be translated into high probability when producing a belief. “[H]owever supportive it may be, statistical evidence will contribute little to the warrant of a conclusion unless it is also reasonably independently secure.”[61]

The focus on knowledge attributions rather than knowledge simpliciter is motivated by the realization that, at trial, truth is always at issue. To say that we ought to be concerned with warranted attributions of knowledge is to say that we are trying to make sure that the beliefs formed by fact finders are sufficiently well-grounded. The view that this sufficient grounding consists of a certain level of probability of truth is simply one way of cashing out what counts as sufficient justification for warranted knowledge attributions.

Michael Pardo points out that truth itself cannot be the focus since some true beliefs can be accidental, and we want evidence law to ensure that fact finders are forming their beliefs for good reasons.[62] This is an additional reason to say that the goal of evidence law is warranted knowledge attributions, rather than truth. By focusing on when an attribution of knowledge is warranted, we are saying that the law is seeking to ensure that fact finders have good reasons for their beliefs about evidence and about what is useful for coming to decisions. Again, even if a preponderance-of-evidence standard is at play and fact finders do not need to attribute to themselves full knowledge of the substance of the verdict in order to reach it, they still need to attribute knowledge to witnesses and to themselves regarding pieces of evidence along the way to the verdict.

Ho Hock Lai notes that when fact finders render a verdict, they are necessarily asserting the truth of the verdict in that the verdict is very simply their finding of fact.[63] He adds that assertions are necessarily claims of knowledge.[64] Among other reasons for this, we cannot assert that something is the case and at the same time claim that we have no knowledge that it is true.[65] Hence all verdicts are claims (or self-attributions) of knowledge. Even where there is a legal presumption in favour of one party and one is finding for that party merely because the opposing party did not sufficiently meet its burden, one is thereby asserting (and therefore generally attributing knowledge to oneself) that the opposing party did not meet its burden. While Ho eventually modifies his view to state that fact finders need only to reach a verdict where they would be justified in believing the truth of the verdict were they to take only admissible evidence into account, even under this view fact finders need to make knowledge attributions to witnesses and themselves as they amass information that would serve to justify their eventual verdict.[66]

[W]hat the finder of fact is asked to determine is not whether the defendant did it, but whether the proposition that the defendant did it is established, to the required degree of proof, by the admissible evidence presented; in other words—subject to the legal constraints signaled by the phrases “to the required degree of proof” and “admissible evidence”—to make an epistemological appraisal.[67]

Furthermore, the focus on knowledge attributions should not be confused with the claim that fact finders must have robust knowledge (rather than beliefs that are highly probably true) in order to reach their verdict. In order to form any beliefs, fact finders will logically be attributing knowledge to witnesses and to themselves, even if the conflicting mass of evidence only leads to a probability that one side’s claims are valid. At the very least, the belief in a higher degree of probability that a given witness is reporting a fact is itself a form of knowledge attribution. To say, “I have a greater than fifty per cent confidence that witness A has reported the facts truthfully” is itself to attribute a kind of knowledge to oneself that the degree of belief is well-founded.[68]

In light of this concern, even an ideal juror (or judge) is best served by certain exclusionary rules. These rules can be justified by appealing to the need for the law to anchor knowledge attributions made by fact finders (to themselves or to witnesses) to a context suggested by the standards of proof appropriate to the case. That is, the standard of proof suggests a certain level of epistemic justification needed for knowledge attributions.

It should be emphasized, however, that in this debate over the exclusion of relevant evidence, I am focusing on rules such as hearsay exclusions,[69] which are not based on external policy considerations, such as the need to incentivize law enforcement officers to avoid unreasonable searches and seizures, or alleged tortfeasors to act swiftly to prevent further accidents,[70] or to protect incentives for people to communicate freely with their doctors, clergy, or lawyers.

While the argument of this paper is primarily occupied with hearsay as an example of an exclusion about which much thinking and rule making has already been done, it is intended to apply more broadly to dissuade those who would curtail epistemically-based exclusionary principles in favour of free proof.

IV. Contextualism

As mentioned above, theorists approach the study of knowledge from the approximation that knowledge is justified true belief,[71] although there are many counterexamples to this formulation.[72] Even within this approximation, questions arise as to what counts as sufficient justification to constitute knowledge. The challenge that theorists attempt to answer with a variety of accounts of knowledge is how to support the notion that a belief is justified where the usual evidence we use to form those beliefs is consistent with the belief being false. That is, if all the evidence we have from our senses and other sources is consistent both with what we believe about the world around us and with the possibility that we are dreaming or hallucinating, then it would seem as though we cannot claim to have any knowledge about the external world. This is the problem of skepticism.[73]

I will focus on epistemic contextualism, one admittedly controversial theory that arose as a reply to the skeptical worry.[74] While it is controversial, there are a number of factors that point in its favour for these purposes. For one, contextualism captures theoretically the suggestion we find in law that there can be different standards for knowledge attributions in different circumstances. The very fact that there are two or three different standards of proof in differing legal contexts suggests, in conjunction with the argument in the previous section, that our legal practices already embrace the notion that the standards of epistemic justification can vary according to context.[75] Hence, if it made any sense to say that the law seems to favour a given epistemic theory (which is not a claim I am making), it would likely favour something like contextualism.

Some competing epistemic views go even further than contextualism in basing the warrantability of knowledge attributions on the practical considerations or interests of those subject to those attributions, or what is properly salient to them.[76] While I will not be able to give a complete picture of those theories here nor a defense of simple contextualism against the alterations and extensions such theories make, suffice it to say that theorists holding these competing views would likely reach similar conclusions about evidentiary exclusion from similar worries. Furthermore, contextualism does so while avoiding the somewhat counterintuitive claim that knowledge itself is directly dependent upon the practical interests at play.[77] Contextualism has the merit of being an approach to knowledge that meets the law’s suggestion of variable epistemic standards, while maintaining certain more traditional intuitions about knowledge and when it is reasonable to ascribe it.

The main thrust of contextualism is that the justification standards for knowledge are contextually bound to the kind of information presented. A contextualist reply to the skeptical hypothesis says that it is only a worry if we are monolithic about the meaning of knowledge. If, on the other hand, the meaning of knowledge changes depending on the context in which it is used or attributed to someone, then we can accurately say that we know, for example, that Alice shot Beth because we saw it, until someone raises a skeptical possibility (for example, that we were hallucinating). At that point, the contextual standards for the attribution of knowledge have been raised and it may no longer be true to say that we know that Alice shot Beth.[78]

David Lewis explained this argument by suggesting that epistemology is “an investigation that destroys its own subject matter.”[79] The possibility of knowledge is forestalled by skeptical hypotheses that arise because of the investigation and testing of what constitutes knowledge. But if this notion of knowledge is correct, it is not only epistemology that destroys knowledge; it is also skeptical scenarios that can be raised in many other contexts, such as the formalized crucible of the courtroom.

As mentioned above, although knowledge is classically understood to be justified true belief, epistemologists today generally agree that such beliefs are neither strictly necessary nor sufficient for knowledge.[80] Nevertheless, justified true beliefs still serve as a good point of departure for discussions about the components of knowledge. That is, one only knows some proposition P when one believes P, P is actually true, and one’s belief in P is justified (i.e., based on good evidence or other good reasons to believe it).[81] In the courtroom, we expect fact finders to make attributions of knowledge to witnesses based on the witnesses’ testimony, and then, on the basis of those attributions, make attributions to themselves, believing themselves to have knowledge (or lack it) in light of what they hear from the witness and other information they get in court, as well as their pre-existing corpus of beliefs.[82] Put somewhat more formally, we generally say that F is justified in believing that P on the basis of W’s assertion that P, only if F is justified in believing that W knows whether P.[83] This places the primary emphasis on the standards by which those knowledge attributions are justified. Exclusionary rules in evidence law can then be justified partially as an attempt to control the context of those attributions.

Contextualism is the theory that the meaning of knowledge in its attribution to someone, as in the sentence “S knows that P,” depends on factors in the context of that attribution which can affect the relevant standards of justification.[84] In other words, the epistemic standards that S must meet vary according to the context in which the trier of fact is uttering (or considering the belief that) “S knows that P.”[85] Contextualism is therefore a theory about the truth conditions of knowledge attributions. When the context is such that the standards are low, it will be easier for knowledge attributions to be true. When the context is such that the standards are high, it will be more difficult. Some of those same knowledge attributions that were true in the low standards context will be false in the high standards one.

Lewis provides a useful set of rules for coming to knowledge understood contextually.[86] While these are admittedly rules of thumb, they do provide support for seeing knowledge attributions as rule-governed, and hence also support the value of having rules to govern the admission of evidence in the courtroom. This also helps to counter Bentham’s claim that ordinary knowledge is not a rule-governed enterprise.

Lewis doesn’t claim that we consciously use these rules in deciding or making our knowledge attributions. Since the rules govern the truth conditions of knowledge attributions, they govern the warrantability of making those attributions. So we are not following those rules consciously in deciding what we believe we know. Rather, the rules govern the correctness of those beliefs.

The courtroom is precisely the kind of place where such rules would need a more formal and perhaps even conscious application. This will be important to keep in mind when we turn to the application of contextualism to the hearsay exclusion; the worry is not (just) that fact finders’ subjective beliefs about knowledge are swayed by listening to hearsay, but that the objective standards of epistemic justification are being manipulated.

This bears repetition: if contextualism is right, then contextual details alter the objective standards for what can count as knowledge. The worry is not about what fact finders do or do not believe, but about the flexibility of objective criteria for assessing their knowledge attributions.

Lewis is answering the question of what possibilities may and may not properly be ignored in making knowledge attributions. That is, when we are trying to decide whether or not to attribute knowledge to a witness (or to ourselves), there are many situations and possible circumstances that would preclude the truth of the proposition believed (or entertained). For example, the possibility that a witness was hallucinating would, if true, undermine the truth of the witness’s claim that he knows Alice shot Beth because he saw it happen. The question is which of these possible circumstances can properly be ignored and which ought to be taken into account for any attribution of knowledge to be true.[87]

Put somewhat more formally, Lewis’s account is that A’s utterance of (or belief in) “S knows that P” is true as long as S’s evidence rules out all possibilities of not-P except those possibilities that are properly ignored by A.[88] Note that the question of what is properly ignored is dependent upon the context that A (the attributor) is in, while it is the evidence held by S (the subject) that is doing the ruling out. Also, Lewis’s notion of evidence here should not be confused with the legal notion, and instead is meant to describe the totality of the subject’s experiential state. Put another way, a fact finder’s attribution of knowledge is appropriate so long as the information to which the witness has access rules out all defeating possibilities, except any possibilities the fact finder may properly ignore.

To apply this in an admittedly oversimplified example, we can imagine a fact finder hearing a witness testifying that he heard a gunshot and immediately burst into the room to see Alice holding a smoking gun over the prone body of Beth. The fact finder can properly attribute the knowledge that Alice shot Beth to the witness. The fact finder thereby comes to attribute that knowledge to herself upon hearing the testimony, assuming she believes the witness and this belief is consistent with the rest of the acceptable evidence that she hears. The attribution of knowledge to the fact finder herself takes place if she is entitled to ignore defeating possibilities that are consistent with the witness’s experiential state: for example, that the witness was hallucinating, or confused someone else with Alice, or that there was a third party who thrust the gun into Alice’s hand in the split second before the witness’s entry.[89]

Here are Lewis’s rules for which possibilities are properly included by the attributor and which are properly ignored: the “Rule of Actuality” holds that anything actually true may not be ignored by the attributor.[90] The “Rule of Belief” holds that anything the subject believes (or that he ought to believe based on the evidence he has) may not be ignored by the attributor.[91] The “Rule of Resemblance” holds that the attributor may not ignore any possibility that is saliently similar to that of the subject’s belief that is the candidate for knowledge.[92] The “Rule of Reliability” is a permissive rule, detailing what may be ignored, rather than what must be included.[93] It holds that we may generally ignore any possibility in which the subject’s usual ways of coming to information fail.[94] Of course, this is very often a subject of doubt in courtroom situations, and this permissive rule is defeated by conflicting cases of rules telling us what may not be ignored. Similarly defeasible, two “Rules of Method” allow the attributor to assume that given samples are representative and that the best explanation of the subject’s evidence is the right one.[95] Another permissive yet defeasible rule is the “Rule of Conservatism”, which allows an attributor to ignore any possibilities usually ignored by everyone else.[96]

The final rule is perhaps the most important one for our purposes: it is the “Rule of Attention”, which holds that any possibility may not be ignored once it has been raised.[97] Once a possibility in which the proposition is false has been brought to the attention of the attributor, then, so long as that context persists, an attribution of knowledge that the subject knows the proposition will be false. This is where the traditional skeptical hypotheses interfere with knowledge. When someone raises the possibility that our senses are deceiving us, then we are in a context where many of our knowledge attributions will be false.

It should also be noted that to the extent that a possibility can be eliminated, either by preventing the possibility from being raised in the first place or by artificially requiring its exclusion from consideration, the contextual standards can be kept low and knowledge can be more easy to come by. While contextual standards are usually kept low by keeping information about possibilities out of consideration, an interesting and neglected corollary of the contextualist view of knowledge is that the contextual standards can also be kept high by excluding certain considerations. If an attributor is in a context of suspicion and doubt, that context can be eased somewhat by relying on the permissive rules to allow the consideration of information, doubts about which may be ignored, where those considerations are not defeated by the other rules. For example, the rule of reliability allows us to rely upon the witness’s usual ways of coming to knowledge and to exclude considerations that those usual ways are misfiring, unless there is information to the contrary.

This leads to a more relaxed context for knowledge attribution. In the example of listening to the witness recount his story about finding Alice with a smoking gun, keeping out information that the witness is very well acquainted with Alice will keep the stakes for knowledge attribution higher by keeping alive the possibility that he confused someone else for Alice. That is, by keeping that information away from the fact finder, it is more difficult to warrant any ascription of knowledge by the fact finder to the witness.

While the literature has not paid much attention to the exact process by which the contextual standards for knowledge attributions are lowered (after all, the epistemologists are primarily concerned with replying to skeptical worries that knowledge is never possible), Lewis suggests that it is as simple as moving the conversation on to another context in which the doubt-inducing possibility is properly ignored.[98] An idea of how this might work can be illustrated by considering another conversational game in which the semantic standards are contextually bound.[99] I finish drinking my glass of water, place the glass on the table and announce: “The glass is now empty.” You point out that several drops remain and I must now agree that the glass isn’t really empty. But then you say: “If it were only those drops in a swimming pool, you’d probably have to say that the pool was empty.” To that, I’d probably agree.[100] You first raised a kind of skeptical consideration for the use of the term “empty”, and then shifted the context to once again make the term easy to use. Now imagine the same scenario, but you never raise the consideration about the swimming pool. In that case, I would be stuck with the higher standards context of admitting that the glass was not empty until something else shifted the context back to a more relaxed one.[101]

Attributions of knowledge are like this but are more complex. In the analogy of the drops of water in the glass and swimming pool, emptiness is analogous to knowledge. The point of the analogy is to explain how context can set the standards for the proper use of the term. Raising the consideration of the swimming pool re-establishes a lower standards context for the use of the word “empty”.

Skeptical considerations can undermine one’s ability to make warranted knowledge attributions. But the activation of what we might call our quotidian epistemic apparatus can shift the context back to one in which the standards for knowledge attribution are lower again. A prime example of a quotidian epistemic apparatus is hearsay. Since we come to so much of our useful knowledge outside the courtroom through hearsay in situations of “natural testimony”,[102] receiving information via hearsay is likely to re-establish a lower standards context for knowledge attributions.

You and I are having a conversation about a hockey game I attended and I say, “I know Smith scored because I saw him hit the puck into the net.” An epistemologist has just entered the room and annoyingly asks, “How do you know you weren’t hallucinating?” Now I ought to agree, “I (believe I) saw it but I don’t really know it because I cannot ignore the possibility that I was hallucinating [any longer].” The conversation turns back to hockey. You tell me that you heard a news report on the radio that our favourite team has been bought and will be moved to another city. As long as the epistemologist stays quiet, it is certainly permissible for me to say, “I know that our team will be moved” (assuming I have no pre-existing reason to doubt you or the news report). Once raised, I cannot properly ignore the possibility that I was hallucinating at the hockey game. But unless the epistemologist starts drawing my attention to that possibility in other contexts as well, I may once again ignore that possibility when considering my belief that the team will be moved.

While the skepticism raised by the epistemologist might extend for a time to similar evidentiary situations, there is no immediate reason to extend it to distinct ways of coming to knowledge unless the skeptical possibility is raised for those ways as well. I might have reason to doubt other things I witnessed at the hockey game, but less reason to doubt information I received aurally much later. Hence, after the epistemologist raises the hallucination possibility about the hockey game, I might not be in a position to claim knowledge for other events I experienced with my eyes around the same time (for example, that I saw two acquaintances fighting in the stands during the game). But the shift in focus to other times, places, and senses by hearing from you about the news report that the team is moving re-establishes a lower standards context, making knowledge attributions more easily true.

It is easy to see why we might want to exclude unfounded skeptical considerations in the courtroom. We do not want merely philosophical possibilities to undermine the basis for knowledge attributions that fact finders make. We want the considerations that might interfere with knowledge attributions to be reasonable doubts (even at the highest standard of proof), and merely philosophical possibilities such as the possibility that I was hallucinating or dreaming about a hockey game are manifestly unreasonable doubts. But at the same time we need to be more attentive to the other side of the contextualist spectrum. The need for a certain heightened epistemic standard in the courtroom provides a reason to avoid the introduction of information that employs our quotidian epistemic apparatus, which tends to depress that heightened epistemic standard.

V. Courtroom

The courtroom is a somewhat artificial epistemic environment: people have an unusually high motivation to lie or dissemble; lawyers are trying to manipulate the information that is brought out so that the situation appears favourable to their side; fact finders are likely to be aware of these concerns and are well advised to be more on guard epistemically than they usually would be. In criminal cases, the standard of proof itself reinforces the need for greater vigilance.

Nevertheless, into this situation people necessarily bring their normal epistemic apparatuses. That is, their means of coming to knowledge are substantially the same as they are outside the courtroom: observation, hearing witness reports, deciding what to believe on the basis of argumentation. The law then needs to manipulate the artificial environment to bring people’s epistemic apparatuses into line with the standards of proof appropriate for the case. This is to ensure that the epistemic methods are appropriate for the heightened standards context of the courtroom. The rules of evidence are tools for doing just that.

There is no precise science of epistemic contexts. Nor is there a bright line distinction between a high standards context and a low standards context. Hence, a legal standard of proof is not suggesting some specific epistemic context that must be maintained throughout the proceedings. Indeed, the fact that the standards of proof are not considered to have sharply defined boundaries[103] may additionally reflect the fact that the epistemic contexts themselves do not have well-defined boundaries. Instead, the standard of proof for the case suggests a relatively narrow range of contexts that exclusionary rules can help to police.

Fact finders are in a high standards context in the courtroom because it is harder to come by knowledge than it usually is in most non-philosophical conversations. That is, fact finders are not entitled to ignore as many knowledge-defeating possibilities as they would be outside the courtroom. Some of this is because inconsistent scenarios are constantly being raised on cross-examination. Additionally, the courtroom is a higher standards context even absent the raising of specific knowledge-defeating possibilities because of the heightened doubt inherent in the context.

Given the ability of skeptical possibilities to introduce contexts where almost all knowledge is impossible, even the heightened standard of proof of a criminal trial suggests a range of contexts where not just any doubt will suffice to defeat knowledge.[104] In Lewis’s terminology, the rule of attention may defeat knowledge attributions as long as the defeating possibility is raised. But the law controls fact finders’ attention by excluding some evidence that would raise the skeptical hypothesis to their attention, and barring that, asking fact finders to exclude certain considerations from their attention (which would be considered impermissible under Lewis’s theory, but can be tolerated in the artificial epistemic environment of the courtroom). Judges may prevent questioning a witness in a manner meant to suggest to the jury that he was hallucinating where there is no other evidence to suggest the witness was prone to hallucinations.[105]

Many exclusionary rules not based on external policy considerations prevent a relaxation of the context away from the range suggested by the standard of proof. If we focus on those exclusionary rules based on epistemic rather than policy considerations (hearsay, character evidence, the control of expert opinion, et cetera, rather than rules designed to deter police misconduct, to maintain the free flow of important communications with lawyers and doctors, or to incentivize remedial measures by accused tortfeasors), the exclusion of such information avoids placing the fact finder in a context in which she has good reason to consider that information in making knowledge attributions. While the epistemologists are usually concerned to show how the exclusion of information can keep lower standards contexts by avoiding the introduction of skeptical worries, some of the information excluded from the courtroom has the opposite (desirable) effect of maintaining a higher standards context.

We can see the best evidence rule[106] as a method for maintaining such higher standards contexts.[107] Whereas in our daily lives, we do not require the best available evidence to have sufficient justifications for our beliefs to make knowledge attributions, in the heightened standards context of the courtroom, we need such rules to maintain that context, requiring higher standards of justifications for those beliefs. The best evidence rule is therefore both a context indicator and context protector for those heightened standards. Hence, relaxations of the best evidence rule (such as those mentioned in the introduction) allowing for the admission of copies of relevant documents and exceptions to the hearsay exclusion weaken the protections of that heightened standards context.

Another example of this can be seen with expert testimony. Given our reliance on experts as the basis of much of our knowledge of the world outside court, hearing expert opinion will tend to entitle a fact finder to ignore more contrary possibilities under the permissive rules of reliability, method, and conservatism. The triggering of these permissive rules is indicative of a more relaxed epistemic context.

I am not suggesting that all expert opinion or hearsay be excluded. Certainly there are things that can be done to maintain a more stringent context even with the introduction of expert evidence, including the ability of opposing counsel to raise doubts about its content. Rather, I am emphasizing an epistemic benefit to these exclusionary rules that advocates of free(r) proof overlook. Again, what is relevant here is not whether the fact finder actually ignores or entertains contrary possibilities that undermine the warrant for knowledge attributions, but that evidentiary rules be designed so as to control when the fact finder has good reason to ignore or attend to them.

Hearsay is the most useful example of how this works. As a primary target of free proof reforms, I will focus on hearsay as a paradigmatic application of the contextualist argument. In the Federal Rules, hearsay is defined as a statement, other than one made by the declarant while testifying at the trial or hearing, offered in evidence to prove the truth of the matter asserted.[108] (Two exceptions that do not count as hearsay are certain prior statements by the witness himself, and certain statements by the opposing party.[109]) In Canada, the language differs slightly in the various cases that treat the issue, but one authoritative statement defines hearsay as

[w]ritten or oral statements, or communicative conduct made by persons otherwise than in testimony at the proceeding in which it is offered ... if such statements or conduct are tendered either as proof of their truth or as proof of assertions implicit therein.[110]

In many less skeptical contexts, second-hand statements are both a reliable and valuable means of coming to knowledge. In the example where you tell me that our favourite hockey team is leaving our city, I don’t usually need to hear it directly from the new owner, or even the news report itself, to attribute the knowledge that the team is moving to you or to myself. But in the courtroom, where such second-hand reports should not suffice for an attribution of knowledge, we want to maintain a higher standards context. Were second-hand statements to be allowed in, a permissive rule like the rule of reliability would allow the fact that it is usually a reliable way of coming to knowledge to relax the context by implying that many doubts about its reliability are properly ignored.

Again, I am not arguing that we exclude all hearsay, just that we should be aware of this epistemic issue when we confront calls for free proof or an elimination of complex exceptions to exclusion in favour of judicial discretion focused primarily on the reliability of the information contained. Also, I am not claiming that all hearsay would create this kind of problem, as hearsay testimony comes in many shapes and sizes. But the rule-based exceptions can be interpreted or adapted to navigate hearsay that does not activate what I have called the fact finder’s quotidian epistemic apparatus.[111]

Admittedly, the judge generally makes a preliminary judgment about the reliability of the information in determining whether to admit the hearsay evidence. Hence, one might think that it is sufficient for epistemic purposes for the judge to make a determination of reliability, which is what the approaches currently in vogue tend to call for.[112] When the judge determines that the hearsay information is sufficiently reliable and allows it at trial, then the judge has determined that there is unlikely to be any significant damage to the fact finder’s ability to try the case using the appropriate standard of proof.

The problem with this rationale is that even if that specific piece of evidence is reliable, its admission helps to decrease the standards for knowledge attribution—not just about that one piece of evidence, but in the wider context of the trial as a whole. Recall that in the hockey example, hearing the report from you about the news item that the team is leaving helps to move us to a lower standards context in which it is easier to make warranted attributions of knowledge, removing the higher standards that the epistemologist imposed with the skeptical worry. That is, while the judge makes admissibility determinations based on the reliability and probative versus prejudicial value of the hearsay evidence, she is not directed to make that determination based on any considerations of what the hearsay evidence does to the epistemic context of the trial as a whole.

The point is not that the information contained in the hearsay itself is suspect, but that by admitting even reliable hearsay, doubts about this and other evidence can more properly be ignored because the fact finder uses a normal, everyday method of knowledge acquisition in assessing the hearsay. The fact finder’s use of that method then makes the epistemic standards for other evidence lower.

It lowers the epistemic standards, not the standards of legal sufficiency or the burden of proof. But the process by which the fact finder reaches a result depends on the epistemic standards as well as the legal ones. Just as it is easier to say correctly that a vessel containing only a few drops is empty when attention is directed at swimming pools, when the fact finder is invited to entertain information obtained through his quotidian epistemic apparatus, other knowledge attributions are more easily warranted.

This objection might be developed and pushed a bit further. It might seem that the impact hearsay has on the epistemic context depends upon how reliable the hearsay is. If the hearsay is independently reliable, in that it is more likely to be conveying true information, then it might seem that attributions of knowledge made on the basis of that hearsay will be more reliable. Hence, so this objection goes, reliable hearsay does not impermissibly lower the contextual standards; it is only the mistaken introduction of unreliable hearsay that does so.

There is a straightforward sense in which it is obviously true that knowledge attributions made directly on the basis of reliable hearsay (attributing knowledge of the content of the hearsay) is better justified when the hearsay is reliable than when it is not reliable. However, this objection misconstrues the relation between the reliability of the hearsay and hearsay’s impact on the wider epistemic context. The ability of hearsay to raise or lower the contextual standards for knowledge attributions does not stem directly from the quality of the hearsay. Reliable hearsay does not necessarily raise the contextual standards and unreliable hearsay does not necessarily lower them. The reliability and content of the hearsay may or may not have a direct impact on the epistemic context.

Recall that the contextual standards depend upon which possibilities inconsistent with the content of the hearsay are properly ignored. Those may not have anything directly to do with the reliability of the hearsay. Reliable hearsay could give rise to more reasons to believe witnesses’ claims by allowing more doubts to be properly ignored, thereby lowering the contextual standards. Or the content might give rise to more reasons to doubt those other claims. The same can be said about unreliable hearsay: it may give rise to more reasons to believe or more reasons to doubt.

This probably seems like I have admitted to the force of the objection. If the content of both reliable and unreliable hearsay can have either a raising or a lowering effect on the contextual standards, then there seems to be nothing to complain about.

The point I am making is that independently of the reliability of the hearsay, there is a way in which its introduction tends to lower the contextual standards for other knowledge attributions by drawing upon the quotidian epistemic apparatus of the fact finder. So long as I have no pre-existing reason to doubt you or the news report about the hockey team moving, even if the report is false, after hearing it my knowledge attributions are more easily warranted.

The distinction to draw here is between the content of the hearsay (and its reliability) and the epistemic package it comes in. In listening to hearsay, the fact finder employs a way of coming to knowledge that works extremely well and often outside the courtroom.[113] This pushes the fact finder into a lower standards context, making other attributions of knowledge more easily justified than would have been the case without the effect of hearsay on the epistemic context.

Let us return to the analogy of the empty water glass and swimming pool and examine it more closely. When you first point out the drops of water left in the glass, you have considerably raised the standards for the use of “empty”, making it harder for claims of emptiness to be true. When you then point out that the same drops in the swimming pool don’t hamper our ability to call it empty, you have then lowered those standards considerably. You might not have lowered them back to the point where it was true when I first claimed the glass was empty. But mentioning the swimming pool has moved the standards back in that direction.

If we were to consider some intermediate container such as a large bucket, it is likely that the first mention of the drops in the glass would have also prevented the truth of an assertion of emptiness for the same drops in a bucket. After calling our attention to the emptiness of the swimming pool containing those drops, however, we are likely in a more warranted position to assert that a bucket with only those drops is empty. This is not something that can be determined with any degree of precision. Rather, the point is that mentioning the swimming pool has moved the context to one of more relaxed standards in which it is easier to be justified in claiming that the bucket is empty than would have been the case after calling our attention to the remaining drops in the glass.

Turning back to knowledge, the presentation of reliable hearsay information admittedly might bring in some very important information that is germane to a correct decision by the fact finder. However, the introduction of this information, coming as it does through the fact finder’s quotidian epistemic apparatus, pushes the fact finder into a context with lower epistemic standards. It is like mentioning the swimming pool when we want to maintain the standard of the drops in the water glass. This is not to say that it is never justified to introduce hearsay, only that we need to be sensitive to the ways in which doing so thereby makes other knowledge attributions more easily justified.

I realize that I have not detailed a sharp distinction between what I am calling the “quotidian epistemic apparatus” and the epistemic mechanisms properly at work in the courtroom. It is clear that there is no fundamental methodological distinction between the ways of coming to knowledge inside the courtroom and the ways we employ every day outside the courtroom. Rather, what does appear to be special about the epistemic method used in the courtroom is that it is performed more self-consciously and on the basis of information the content and presentation of which tend to be outside usual daily experience. The advantage of excluding hearsay and of other epistemic exclusionary rules is that they help to maintain the rarefied epistemic context we need for the courtroom. The claim is therefore that evidence which activates an epistemic methodology more characteristic of the fact finder’s daily experience depresses the epistemic standards at play where we have good reason to maintain a higher standards context.

In fashioning and applying evidence law, we need to be sensitive to the impact of evidence on the standards of knowledge attributions made in each stage of the trial, including (especially) the attributions made when reaching a verdict in the final stages of the decision-making process. Sometimes the propositions contained within some hearsay evidence are true and are useful to reaching a correct verdict. But most often, the information to be presented via hearsay, even if fully credible, is not itself sufficient to warrant a given verdict. The point is not about the relation of the hearsay evidence to verdict itself. Rather, the fact that those propositions come in the form of hearsay threatens the higher contextual standards of the trial generally.

Hearsay, even about non-ultimate facts, can change the salience of defeating considerations, making them appropriately dismissed more easily. In other words, getting true information that is justifiably believed through hearsay may make it easier to justify belief in other propositions considered at trial. When belief in those other propositions is more easily justified, the standards of the trial as a whole are depressed, making it too easy to justify beliefs in propositions of ultimate importance to the verdict, even if those other propositions are not themselves presented in the form of hearsay.

VI. Objections, Replies, and Applications

A Benthamite might respond to these considerations by noting that we could just bring any doubts to the fact finder’s attention to maintain a high standards context. That is, we could simply raise the standards again after the introduction of hearsay by reemphasizing the many reasons to be skeptical about hearsay information.[114] A slightly less Benthamite version of this objection would recall the usual justification for the exclusion of hearsay: the inability to cross-examine the declarant (or, more generally, the inability of the fact finder to assess the credibility of the declarant).[115] The various exceptions to the exclusionary rule and the principled approach are designed to assuage fears about the value of proffered hearsay by pointing to indicia of reliability (where present), thereby overcoming concerns about confrontation and the credibility of the declarant.[116] As long as those fears are dealt with, any surviving hearsay is properly admitted when necessary and reliable under the principled approach.[117] Hence, the high standards context can be maintained by instructions to the fact finder or by the process of vetting the hearsay.

There are a number of problems with this reply. First (and this point bears repetition), the damage that is being done by the introduction of hearsay is not necessarily done by the content of the hearsay itself but the fact that the package in which it comes draws the fact finder into a lower standards context. Doubts about that particular information may not be sufficient because the fact finder is still invited to assess the information using her quotidian epistemic apparatus.

Recalling the example where you report the hockey team is moving, if I were to learn at the same time that the news reporter you heard had a history of making up sensational stories about sports teams, I may come to doubt the information about the team’s moving. But I am not then in a context requiring more reservations about my knowledge attributions on the basis of subsequent hearsay (for example, when you tell me what you heard in the weather report). The only way to raise the standards again successfully for the trial as a whole would be to raise a skeptical possibility about the fact finder’s epistemic abilities or to otherwise shift them back into a context of greater doubt generally. So the current process used to determine the admissibility of hearsay will not protect a higher standards context since it does nothing to keep the fact finders in a context of greater doubt or otherwise introduce any skepticism about their epistemic process.

Second, this is where the law’s need for consistency comes in. If we are trying to maintain a context that is anchored to the standard of proof, it would not be appropriate to let that context fluctuate rather rapidly (perhaps even statement to statement during the testimony of the same witness) by allowing more fanciful skeptical doubts to be raised about hearsay, but not allowing them to be raised about the witnesses’ own experiences.[118]

In the courtroom, there are two knowledge attributions that matter to us, both made by the fact finder. One is when the fact finder is attributing knowledge to a witness. The other is when the fact finder is attributing knowledge to herself. When the fact finder is attributing knowledge to a witness, she is determining that the witness’s experiences comport with what he is reporting and that those experiences justify the witness’s belief in what he is reporting. To do so accurately, the fact finder must include any possibilities she herself must consider under the epistemic rules above.

When the fact finder is attributing knowledge to herself on the basis of the witness’s reports (or documentary evidence), she must also include any considerations raised by the parties and the directions of the judge (if distinct from the fact finder). This includes making determinations about whether the witness was dissembling or had some impairment of his senses.

I do not mean to suggest that knowledge attributions to the witness and to the fact finders themselves are immediately and easily separable. Most often, the two attributions come together unless the fact finder has any reason to consider doubts about her own perceptions of what the witness reported, or when she has reason to doubt the witness’s access to the information presented but then gets corroboration of that information. The fact finder also must make further attributions of knowledge to herself based upon the accumulation of evidence and considerations raised by the lawyers and the judge.[119] As far as the rules of evidence are concerned, it is always the circumstances of the fact finder as attributor that need to be controlled vis-à-vis which possibilities are open or closed to consideration in making knowledge attributions.

It is clearly the case that reliable hearsay increases the probability that knowledge attributions made about the propositions contained within that hearsay are true, and that those knowledge attributions are therefore better justified than if made without that evidence. The problem is in limiting the consideration of whether to admit this evidence by focusing exclusively on the reliability of that piece of evidence, without considering the wider impact on epistemic standards. This is not to say that hearsay should be entirely excluded. Rather, it is to resist the claim that having the court assess its admissibility on the basis of reliability alone is sufficient (and to resist more forcefully calls for free proof). Hence, simply pointing out the many cases in which hearsay is useful in coming to an outcome that is deemed correct is not an answer to this complaint.

One might say that the rules of evidence are more about truth than about knowledge. Both the United States Federal Rules and various Canadian cases specifically state that the purpose of the admissibility rules is to enhance and ensure the reliability of information presented, or to help fact finders decide whether witnesses are telling the truth.[120] Indeed, much of the fact finders’ decision making is taken up with deciding what to believe, using various indicia of truthfulness and reliability for the witnesses from whom they hear testimony. But this is not really a criticism. In deciding what to believe, fact finders are deciding what witnesses actually know, where the truth of what witnesses report is affected by their motivations and propensity for dissembling, and by their reliability at coming to the knowledge they profess to have. Fact finders are therefore making knowledge attributions to witnesses, and then to themselves on the basis of what they hear from witnesses.

The fact that evidence rules may only occasionally employ the term “knowledge” does not mean that they are not still dealing with the conditions of knowledge attributions. In navigating our daily lives, we are constantly making knowledge attributions to others and to ourselves in determining what to believe and only occasionally do we use the term. In the courtroom, we have a collective interest in ensuring that the standards by which knowledge is attributed are linked to the standards of proof the law demands for the case. We want criminal convictions to take place only when the fact finder can attribute knowledge of guilt to herself beyond a reasonable doubt.

In civil cases, we want results to be based on knowledge attributions that reflect the preponderance of the evidence. This might allow for some results not to be based on knowledge, strictly speaking, since one might believe the preponderance of the evidence favours one result but not attribute knowledge to oneself in reaching that result. But even that only threatens the knowledge basis of the outcome itself. Fact finders are still making plenty of knowledge attributions in determining what to believe by a preponderance of the evidence. That is, each piece of evidence in determining what the preponderance yields itself involves a knowledge attribution. Hence, the rules of evidence still take knowledge as the baseline of what fact finders are trying to reach; therefore, the exclusionary rules are still an attempt to bind the context of any knowledge attributions to the particular epistemic environment of the courtroom.

When we step back and consider the wider fairness of our criminal justice system, the epistemic problem with freer admissibility of even reliable hearsay becomes starker. Since not all cases will involve the introduction of hearsay evidence, there will tend to be differing contextual standards of justification for knowledge attributions in criminal cases that do involve the admission of hearsay from those that do not. Trials that involve the admission of information about which knowledge attributions are made using what I have called our quotidian epistemic apparatus will have lower standards for many knowledge attributions involved in determining outcomes. In essence, it is easier to make warranted knowledge attributions in those cases than in cases where such information is excluded or not present. One effect of this is that trials involving hearsay and other such information, covered by epistemic evidence rules, involve different standards of justification for knowledge attributions than trials that do not. The very nature of the evidence presented at trial affects what doubts (possibilities) are reasonable to consider.

An implication of this discussion is that what counts as a reasonable doubt depends on the context. What is properly ignored under Lewis’s rules, and under contextualism more generally, partially determines which doubts are reasonable. While the possibility that the witness and fact finder are brains in a vat may never be a reasonable doubt, if the evidentiary rules and circumstances of the case allow a real possibility that a witness was hallucinating, then raising that possibility to the attention of the fact finder likely creates a reasonable doubt where the witness is giving the only evidence of a material fact.

If I am right that the admission of hearsay lowers the epistemic context, then where it is admitted, fewer possibilities to defeat knowledge attributions will be reasonable, and more can properly be ignored unless raised to attention. Criminal defendants are therefore not being tried according to a uniform standard of proof beyond a reasonable doubt since the reasonableness of the doubt is dependent on the context. Admittedly, this is already an imprecise standard; given the differences in evidence presented, judges, and fact finders, there is necessarily going to be considerable variation in what constitutes the reasonable doubt standard. However, contextualism about knowledge shows that the admission of hearsay makes it even more elastic, where we are best served trying to minimize that elasticity.

One might note that all of a prosecutor’s evidence will tend to make it more difficult for knowledge attributions about facts in support of the defense’s case to be justified, so there is nothing special about these epistemic considerations. But the point here is not about the crucible of duelling factual claims inherent in the opposing sides’ cases. It is very likely that every piece of information presented by one side will tend to undermine the story being told by the opposing side. Rather, the introduction of any evidence that requires the fact finder’s quotidian epistemic apparatus will tend to lower the otherwise heightened standards called for by the standard of proof. The context in which we use the quotidian epistemic apparatus (that is, the more familiar epistemic circumstances of daily life) is one in which more doubts are properly ignored. Hence, drawing the fact finder into that context makes it harder for doubts to be reasonable (in that more doubts are properly ignored). While making it easier to dismiss or ignore doubts could be used against either side, in a criminal trial it is more of a detriment to the defendant.

One might note that the admission of hearsay can just as easily be used against the prosecution as against the defendant. An easy reply to this is that the hope of more uniform standards for criminal trials should be seen as valuable for the prosecution as well as for the defense. This might, however, lead to a concern that defendants are being denied access to valuable exculpatory evidence.

There are two replies to this concern, which cut in opposite directions dependent on one’s views of whether it is more just to focus on the uniformity of standards or on the preferential treatment of defendants. One reply would be to say that since the law is trying to pin a context for the meaning of knowledge to the “beyond a reasonable doubt” standard found in the criminal trial, the admission of hearsay by either party jeopardizes the context’s justificatory standards, and it should not be admitted unless there is some way of maintaining the high standards context. That is, hearsay would be more acceptable if there were some way to ensure that it would not present itself as hearsay to the fact finder such that she would revert to her quotidian epistemic apparatus, instead of maintaining the higher contextual standard.

Those sympathetic to this reply might countenance proposals to couch the admission of hearsay evidence with extra cautions by the judge, or perhaps more radically, with demands that witnesses include some warning in their testimony about the reliability of the hearsay evidence. Otherwise, on this view, hearsay should not generally be admitted for fear of allowing the contextual standard to float too freely, while exceptions should be limited, detailed, and rule based where those rules are sensitive to the impact of the kind of hearsay on the epistemic context.

The other possibility would be to allow hearsay if presented by defendants, allowing them to parse for themselves whether the advantages of the hearsay information outweigh the relaxed epistemic standards and sacrifices to the uniformity of those standards across trials. This would give defendants the option of relaxing the epistemic standards if they believed that the information they can get in by doing so will be helpful to their case. It would sacrifice greater uniformity in epistemic standards across trials for the sake of allowing defendants to present potentially exculpatory evidence.

In admitting hearsay and relaxing the epistemic standards of their trials, defendants would be making more actual or potential doubts unreasonable and properly ignored by fact finders. Recalling Lewis’s permissive rules, discussed above, fact finders may ignore doubts that suggest the witness’s normal ways of coming to knowledge are misfiring or that there are other explanations for the witness’s experiences than the content of his report. When those kinds of doubts are raised to fact finders as live possibilities, they are then reasonable doubts (both legally and epistemically). When hearsay is introduced and the epistemic standards are thereby depressed, more of those doubts will again be properly ignored and hence unreasonable. But perhaps this is the gamble we should leave in defendants’ hands.

Either way, all other considerations being equal, greater uniformity of epistemic contexts for criminal trials would clearly tend to be more fair than allowing these contexts to float with only reliability and necessity to guide admissibility.

These more practical considerations might give rise to another objection:[121] that it is not the role (or expertise) of the court to make determinations regarding the epistemic context of the trial, and that these considerations are simply too rarefied for treatment by judges. While it is certainly doubtless that we cannot expect our judges to appreciate and apply the implications of complex epistemological arguments alongside their existing arduous tasks, we can expect them to have a heightened sensitivity to epistemic contexts if such were mandated by statutory rules or controlling precedent. That is, the suggestions outlined here do not require as much philosophical subtlety as did the argument in their favour. It is enough initially to demand that those arguing for an increase in exceptions or a relaxation of exclusionary rules address the epistemic implications of their policy recommendations before courts or legislatures decide to enact them.

Where such changes have already been made, these considerations should be seen as arguing for a return to a more rule-based exclusionary regime as against one that increases judicial discretion regarding the admissibility of hearsay and similarly problematic evidence. In other words, the objection can be turned on its head: if these epistemic considerations are valid, the very inability of judges to apply them is an argument in favour of reduced discretion.

Furthermore, apart from the rationale for this sensitivity, I am not sure that judges are so incapable of deploying sensitivity to epistemic context when and where judicial discretion is unavoidable. While the finer points of epistemic contextualism may be beyond their purview, it would be enough for judges to be mindful of the implications that admitted evidence has on shifting justificatory standards. We are already used to the fact that the “beyond a reasonable doubt” standard (for example) shifts from trial to trial, and that it is part of a judge’s job to minimize this movement by giving guidance to fact finders. If we merely point out that the kind of evidence admitted at trial can itself have an impact on which doubts are reasonable, judges may see that as an additional reason in favour of excluding evidence that makes more doubts unreasonable.

One might raise a more philosophical objection here. Since the contextualist point is that the meaning of knowledge changes with the context, it would seem that there could be no cause for complaint. Knowledge is simply easier to come by in cases where hearsay is presented. No one should complain about that, since the only criterion we should be concerned with is that cases are decided on the basis of warranted attributions of knowledge. Where it is easier to make such a warranted attribution, the meaning of knowledge is relaxed, but it is still the case that in every criminal trial the standard is knowledge “beyond a reasonable doubt”. Even if the meaning of knowledge shifts, since the standard is tied to that meaning, we cannot complain when the standard is affected by that shift.

The reply to this objection starts with the realization that the rules of evidence function as a bulwark against the variable and shifting contexts in which we make knowledge attributions outside the courtroom. While much of evidence scholarship is focused on the use of legal evidence rules in guaranteeing the reliability of information presented in court, the ultimate point of that focus is the need for fact finders to have good reasons to believe in their determinations of what is true and false at trial. While not detailing the form that epistemic justifications must take,[122] what contextualism shows us is that there is more to the notion of a good epistemic justification than simply the reliability of the information.

Just as we would not want the outcome of a trial to be dependent upon the notion that a glass with a few drops left is not empty, while another trial is dependent upon a swimming pool with a only few drops left being properly considered empty, we do not want some criminal trials to be based on one set of what constitutes good reasons to believe, while others are based on a radically different set of what makes for good reasons. While it is true, according to contextualism, that the meaning of knowledge changes according to the context, what brings about that change is the shifting standard of epistemic justification from context to context. The fact that criminal law gives us a specific (although vaguely defined) standard of proof indicates that the law is seeking to limit those shifts. Since the rules of evidence are understood in terms of their relation to that standard of proof, we can understand them as an attempt to prevent shifts in the epistemic justificatory context in addition to the traditional understanding of them in terms of their control on the reliability of the information presented at trial.

Another objection notes that since every piece of information has a slightly different context from every other, focusing on hearsay as a class to be excluded is useless. The context will shift rapidly even with hearsay evidence excluded: even eyewitness statements will present different contexts from documents, and these will differ from expert testimony about physical evidence.[123] Even within each of these types of evidence, the contexts will differ greatly from one piece to another. Given this huge diversity of contexts, the legal system cannot hope to make bright line distinctions nor treat criminal defendants equally, as the kinds of evidence available in each case will be vastly different.

This is, of course, absolutely true. But we can still make a distinction between those kinds of information that are evaluated differently in the courtroom and those that call up only what I have called our quotidian epistemic apparatus. While the task of differentiating types and contexts of evidence may be too much for the legal system to hope to handle exhaustively, this does not mean that it should not still seek to minimize the impact of those varying contexts on criminal verdicts. We cannot hope to promise criminal defendants that they will be tried using a perfectly uniform set of epistemic justifications, but we can at least offer them a slightly less chaotic set of contextual standards. Consistency in epistemic standards is an ideal that we cannot hope to achieve perfectly. Yet that fact is not a reason not to maximize that consistency to whatever extent possible.

I would be remiss if I did not distinguish this argument from that of two evidence theorists who advance arguments against hearsay on the basis of epistemological claims. Ho Hock Lai argues that the introduction of some hearsay should be restricted out of considerations of justice. However, his focus is on the moral considerations owed to the fact finder in her epistemic process[124] rather than the implications that the epistemic context has for criminal defendants.

Ho also argues that we must separate the rule against the using of or reasoning based on hearsay evidence from the technique of enforcing it, which is generally exclusion.[125] While I agree with this for the most part, it leaves out an important piece. Where the out-of-court statement is not being offered for the content of the evidence but for other purposes (for example, fear in the caller’s voice[126]), and if it is possible to isolate the jury from the content and have them focus merely on the context, then that is not necessarily requiring their quotidian epistemic apparatus and depressing the contextual standards. But at the same time, seen from the standpoint of epistemic contextualism, the technique of enforcing the hearsay rule (exclusion) is not as isolated from the content of the rule as Ho envisions it. If the danger of hearsay is the lowering of epistemic standards, then we are not only concerned with how the fact finder will use or reason on the basis of hearsay. We are worried about what the hearsay will do to the epistemic context of the trial. Hence the exclusion is itself an element of maintaining that high standards context.

True, to be hearsay, the evidence must be offered to support the truth of the matter asserted therein, and it is the focus on that usage of the testimony that creates the epistemic problem. Evidence that would otherwise be hearsay may not require the quotidian epistemic apparatus where the use of the testimony (assuming this can be isolated from its content) is to show, for example, the emotional state of the declarant or her ancillary beliefs. To access those aspects of the testimony, a fact finder probably needs to have her attention specifically directed at them, and so it is not employing the quotidian epistemic apparatus.

The other argument is that of Craig Callen, who uses the linguistic theories of Paul Grice to argue that hearsay interferes with the fact finder’s ability to assess properly the conversational implicature or communicative intention of the declarant’s statement.[127] Callen is partially concerned with using Grice’s theory to delineate more clearly between hearsay and non-hearsay.[128] The main argument appeals to the need for fact finders to employ their everyday epistemic abilities in reaching their conclusions.[129]

While I agree with the general push to provide a firm basis for hearsay exclusions, my argument differs from Callen’s in several key places. For one, my focus is more directly on epistemology while his is on the possibilities and norms of communication, both generally and between the declarant and fact finder. Furthermore, while Callen is investigating those norms of communication from the standpoint of fact finders’ normal epistemic and communicative abilities, I am emphasizing the need for the court to project a rarefied epistemic air so as to maintain a higher standards context, and specifically to avoid depressing that context by appealing to those normal epistemic abilities.


The rules of evidence are about knowledge, of which truth is a part. But truths can only be useful to a trier of fact where she also has good reason to believe those truths. Hence, the rules of evidence must be about determining the conditions under which information is to be displayed in court so that triers of fact can have good bases for their beliefs in that information.

Contextualism is simply a philosophical elaboration upon the basic point that what counts as good reason for belief varies depending on the context in which the information is considered. The courtroom is a special context that needs to maintain that special quality. So epistemic rules of evidentiary admissibility serve to maintain the quality of that special context and fix what counts as good reason to believe the information presented. The admission of hearsay puts the fact finder in a context in which it is easier for her to have good reasons to make attributions of knowledge. This means that knowledge attributions are more easily made true in those cases, and that criminal defendants are not being treated uniformly by the process.