II. A COHERENCE THEORY OF EVIDENCE AND LEGAL PROOF

The main tenet of the coherence-based theory of evidence and legal proof that I argue for may be succinctly stated as follows:

Some notions need to be clarified before we are in a position to unpack this claim.

III. FACTUAL COHERENCE AS CONSTRAINT SATISFACTION

The first element that is essential to a viable coherence theory of evidence and legal proof is a concept of coherence. A conception of coherence as constraint satisfaction developed by Paul Thagard provides a suitable framework for analyzing the kind of coherence that is relevant to the justification of evidentiary judgments in law. According to this conception, coherence is a matter of the satisfaction of a set of positive and negative constraints among a given set of elements. In order to achieve coherence, we divide up a set E of elements, which may be propositions or other representations, into two disjoint subsets—A, which contains accepted elements; and R, which contains rejected elements—by taking into account the coherence and incoherence relations that hold between pairs of elements of E. For example, if a hypothesis h1 explains e1, we want to ensure that if h1 is accepted, so is e1. On the other hand, if h1 is inconsistent with h2, then we will make sure that if h1 is accepted, then h2 is rejected. On this view, coherence results from dividing up E into A and R in a way that best satisfies the most positive (coherence relations) and negative (incoherence relations) constraints.Footnote ¹¹

In order to apply this approach to coherence to a particular problem, we need to specify the elements and the constraints that are relevant in a particular domain as well as the kinds of coherence involved. Thagard distinguishes six kinds of coherence, namely, explanatory, analogical, deductive, perceptual, conceptual, and deliberative. Each of these kinds of coherence is specified by means of a set of principles that state the relevant elements and constraints. In the case of explanatory coherence, which is—as I argue later—the most important contributor to the justification of evidentiary judgments in law, the principles read as follows:Footnote ¹²

• Principle E1: Symmetry. Explanatory coherence is a symmetrical relation, unlike, say, conditional probability. That is, two propositions p and q cohere with each other equally.

• Principle E2: Explanation. 1) A hypothesis coheres with what it explains, which can be either evidence or another hypothesis; 2) hypotheses that together explain some other proposition cohere with each other; and 3) the more hypotheses it takes to explain something, the lower the degree of coherence.Footnote ¹³

• Principle E3: Analogy. Similar hypotheses that explain similar pieces of evidence cohere.

• Principle E4: Data Priority. Propositions that describe the results of observation have a degree of acceptability on their own.

• Principle E5: Contradiction. Contradictory propositions are incoherent with each other.

• Principle E6: Competition. If p and q both explain a proposition and if p and q are not explanatorily connected, then p and q are incoherent with each other (p and q are explanatorily connected if one explains the other or if together they explain something).

• Principle E7: Acceptance. The acceptability of a proposition in a system of propositions depends on its coherence with them.

Thus explanatory coherence is a symmetrical relation between hypotheses and evidence within a set (E1). Explanatory coherence arises out of relations of explanation and analogical relations between evidence and hypotheses (E2, E3). Relations of contradiction and competition give rise to incoherence (E5, E6). And the acceptability of a proposition is claimed to be a matter of its coherence with the rest of propositions within a set (E7), some of which enjoy, nonetheless, a degree of acceptability on their own (E4).Footnote ¹⁴

Let me illustrate these principles by means of a legal example, the Regina v. Angela Cannings case.Footnote ¹⁵ Angela Cannings was the mother of four children, Gemma, Jason, Jade, and Matthew. Three of these children (Gemma, Jason, and Matthew) died in infancy. She was charged with the murder of both of her sons, Jason and Matthew. The charge for the murder of her first child, Gemma, did not proceed. Three of the children, including the daughter who survived, Jade, had suffered an acute or apparent life-threatening event (“ALTE”). All the ALTEs and the deaths occurred when the babies were in sole care of the mother. The prosecution alleged that the defendant had smothered both her sons, intending to kill them by obstructing their upper airways. To support that allegation it was suggested that the death of Gemma and each of the ALTEs suffered by the other children were also the result of smothering by the defendant and that these actions formed part of an “overall pattern.”

The defendant, a woman described by hospital staff as a loving mother, denied harming any of her children. Her case was that the deaths were natural, if unexplained, incidents to be classified as sudden infant death syndrome (SIDS). The Crown adduced expert evidence as to the extreme rarity of three infant SIDS in one family, which suggested that smothering was the most likely diagnosis. There was no medical evidence that Mrs. Cannings had any personality disorder or psychiatric condition. Nevertheless, as was accepted without hesitation by Dr. Johnson, parents who appear to be affectionate and caring toward their children do sometimes kill them. Expert medical witnesses on behalf of the defendant supported the claim that three unexplained infant deaths in the same family do not, however, inexorably lead to the conclusion that they must have resulted from unnatural causes.

On April 16, 2002 in the Crown Court at Winchester, the defendant was convicted of murdering both her sons. On appeal the convictions were quashed. Significant new evidence was presented before the court. After the trial, further investigation was carried out into the extended family. This investigation revealed further infant deaths. One of these deaths was particularly relevant to the case. Unbeknownst to the defendant at trial, she had a half-sister, who following these convictions made herself known to the defendant's attorneys. She had three children, two of whom had suffered ALTEs. This, according to the court, supported the view that there might well be a genetic cause for the deaths of the Cannings children. A substantial body of medical research, not put before the jury but received by the Court of Appeals in evidence, further fortified the analysis that there was a reasonable possibility that the deaths of these children could be natural, even if unexplained by the present state of knowledge. If, to the contrary, the appellant had indeed killed her children—as the jury found that she had—the logical conclusion about her repeated pregnancies was that she was having babies in order to kill them. That, said the court, predicated an extraordinary state of mind contradicted by the uncontested evidence about the love and care Mrs. Cannings bestowed upon her children.

The hypothesis that Angela Cannings killed her babies explains the evidence that the babies are dead, so the hypothesis and the evidence cohere with each other, as Principle E2, Explanation, states. Although the relation between the explanans and the explanandum is asymmetrical, with hypotheses explaining evidence but not vice versa, the coherence relation between them is symmetrical, as Principle E1, Symmetry, establishes. Thus, if the hypothesis that Angela Cannings killed her babies coheres with the evidence that the babies are dead, then that evidence and the guilt hypothesis also cohere.

Principle E2 also allows the possibility of hypotheses explaining each other, as when the hypothesis that Angela Cannings was the murderer is explained by the hypothesis that she was only apparently a loving mother. This principle also says that hypotheses that together explain some other proposition cohere with each other. For example, the hypothesis that the testimony of hospital staff that there was no evidence of ill-treatment or lack of care and the hypothesis that there was a genetic condition in the Cannings family together explain the evidence that the babies showed no sign of physical interference; thus these hypotheses cohere with each other. The last part of Principle E2 says that the more hypotheses it takes to explain something, the lower the degree of coherence. For example, the hypothesis that Angela Cannings orchestrated but did not commit the murders requires multiple hypotheses that lack simplicity.

Principle E3, Analogy, states that similar hypotheses that explain similar pieces of evidence cohere. For example, if Angela Cannings had a history of violent acts toward her children, then these cases provide analogies that make the hypothesis that the deaths of the babies resulted from deliberate infliction of harm more plausible.

The first part of Principle E4, Data Priority, says that propositions that describe the results of observations, for example, the absence of any evidence of physical violence in the postmortem medical examinations, have a degree of acceptability on their own; that is, that they have a priority in being accepted. According to Principle E5, Contradictory propositions for example, the hypothesis that the infant deaths were natural and the hypothesis that they were unnatural are incoherent with each other.

Incoherence relations may also be established between two hypotheses if they are in competition—as Principle E6, Competition, says. Two hypotheses compete with each other if they both explain a proposition but are not explanatorily connected. For instance, the hypothesis that Jason and Matthew had a genetic defect and the hypothesis that they were killed compete with each other even though these are not contradictory, for it is logically possible both that the children had a genetic defect and that they were smothered. Since both hypotheses explain the evidence of the infant deaths, but neither one explains the other, nor do they together explain the evidence, they are incoherent with each other.

Last, Principle E7, Acceptance, says that the acceptability of any proposition (e.g., that Mrs. Cannings killed her babies) depends on its coherence with the rest of the elements to which it belongs.

A theory of “factual coherence,” that is, the kind of coherence that is relevant to the justification of evidentiary judgments in law, may be developed along the lines of Thagard's coherentist account of epistemic justification. Nonetheless, several modifications at all three levels—that is, the kinds of coherence involved, the constraints, and the relevant elements—are necessary to make Thagard's theory of epistemic justification applicable to the justification of beliefs in the legal context.

First, Thagard claims that epistemic justification involves the interaction of deductive, analogical, perceptual, conceptual, and explanatory coherence. This conception of epistemic justification agrees with the received tradition in epistemology, according to which epistemic justification has to do exclusively with our truth-related goals. This “intellectualist” conception of justification is, however, problematic.Footnote ¹⁶ As has been persuasively argued in the last decade, the epistemic justification of a belief does not depend only on epistemic factors but also depends, at least partly, on pragmatic considerations.Footnote ¹⁷ Our practical goals and concerns also have a bearing on whether or not a given belief is epistemically justified: two subjects S and Sˊ might be alike with respect to the evidence they have in favor/against p, but because of differences in their pragmatic situation (e.g., the stakes), whereas S is justified in believing p, Sˊ is not.Footnote ¹⁸ That is to say, a subject's pragmatic situation may affect his justification.Footnote ¹⁹

The justification of beliefs about the events being litigated at trial is similarly sensitive to variations in pragmatic factors.Footnote ²⁰ Legal fact finders often take (and ought to take) practical factors into account when forming justified beliefs about disputed propositions at trial.Footnote ²¹ In order to account for the “pragmatic encroachment” on justification, I suggest that we modify Thagard's theory of epistemic coherence by adding “deliberative coherence,” that is, the kind of coherence that is relevant to practical inference.Footnote ²² By bringing practical considerations into the computation of coherence, we thus give pragmatic factors their due in the justification of a fact finder's beliefs about the events being litigated.Footnote ²³

Second, legal reasoning is a highly institutionalized kind of reasoning. Legal reasoning about evidence takes place within a complex institutional framework. The evaluation of hypotheses about facts at trial is carried out within a normative framework that requires a skeptical stance toward the guilt hypothesis: this hypothesis may be accepted only if belief in this hypothesis is beyond a reasonable doubt and provided that the presumption of innocence is rebutted. The reasonable doubt standard and the presumption of innocence give rise to “institutional constraints,” which need to be satisfied when maximizing coherence. Specifically, the presumption of innocence may be treated as a constraint that requires that hypotheses compatible with innocence be given priority in being accepted, and the reasonable doubt standard requires that the guilt hypothesis be accepted only if its degree of justification is sufficiently high to meet the standard.Footnote ²⁴

Third, the institutional dimension of evidential reasoning in law also puts some limits on the kind of elements that may be included within a coherence calculation. Factual accuracy is surely a momentous value of legal systems. However, the legal system is meant to serve values other than truth. Rules of evidence and procedure are in place to ensure that the complex set of values that the law aims at advancing is protected. These norms make inadmissible evidence that would be deemed relevant were we able to pursue inquiry free from institutional boundaries. Thus only some subset of the evidence available (i.e., the admissible evidence) enjoys the kind of data priority that the principles of explanatory coherence grant.Footnote ²⁵

In sum, in order to apply Thagard's coherence-based theory of epistemic justification to law, it is necessary to add new constraints, that is, institutional constraints, to modify some of the constraints proposed within the theory, for example, the principle of data priority, in order to limit the set of elements that may enter into the coherence calculation, as well as to include another kind of coherence, that is, deliberative coherence. The conception of factual coherence that results from these modifications may be stated as follows:

Each of the foregoing kinds of coherence is relevant to the justification of factual propositions in law. “Analogical coherence” plays a prominent role indeed in fact-reasoning in law. As empirical research shows, some kinds of analogy are highly relevant to jurors’ reasoning processes, such as the analogy to the self, for example, “if somebody threatened me in a pub, I would not go back to that pub unless I were looking for revenge; therefore, the defendant came back to the pub looking for revenge.”Footnote ²⁶ “Conceptual coherence” is also pivotal in fact-reasoning in law. This kind of coherence is critical when applying stereotypes, which is, arguably, an important part of reasoning about evidence in law.Footnote ²⁷

Although reasoning about evidence is certainly nondeductive, deduction has surely a role to play when drawing conclusions about disputed questions of fact. “Deductive coherence” is thus at the very least one of the kinds of coherence that are relevant for the justification of factual conclusions in law. “Perceptual coherence” is required to give an account of the fact finder's reasoning processes that involve nonverbal representations. Some of the factual inferences in law are best understood as “visual” abductions, that is, abductions that use pictorial or other iconic representations.Footnote ²⁸ For instance, a picture of the scene of a crime may suggest a likely narrative about how the events took place. “Deliberative coherence” plays a role in determining which decision about the events at trial is best in light of the several goals that the law is meant to serve. Last, “explanatory coherence” is the most important kind of coherence in a theory of the justification of evidentiary judgments in law. As argued, some of the principles of explanatory coherence need to be modified so as to account for some features specific to factual reasoning in law. More precisely, Principles E4 and E7 should be revised to read as follows:

• Principle E4ˊ: Data Priority. a) Propositions that describe admissible evidence at trial have a degree of acceptability on their own; b) factual hypotheses that are compatible with innocence have a degree of acceptability on their own.

• Principle E7ˊ: Acceptance. a) The acceptability of a proposition in a system of propositions (i.e., a theory of the case) depends on its coherence with them; b) the guilt hypothesis may be accepted only if it is justified to a degree sufficient to satisfy the reasonable doubt standard.

The revised version of Principle E4 is both narrower and broader than the original one. It is narrower, as only some of the propositions that describe the results of observations, that is, those that describe admissible evidence at trial, enjoy a degree of priority. And it is also broader, for it is not only evidence propositions that should be given priority in the legal context, but hypotheses compatible with innocence also have a degree of acceptability on their own. Principle E7ˊ significantly restricts the propositions that may be accepted on grounds of their coherence to those that meet the applicable standard of proof.Footnote ²⁹

To sum up, the constraint-satisfaction approach to coherence provides a useful framework for developing a conception of coherence that is usable in the context of legal fact-finding. Some modifications are necessary in order to account for the specificities of evidential reasoning in law, most importantly, the fact that evidential reasoning in law is practical and institutional in nature. These modifications, I believe, bring the model of coherence as constraint satisfaction closer to capturing the complexities of legal reasoning about facts.Footnote ³⁰ Let us now examine the issue of how factual coherence defined in this way may be established in the course of legal decision making.

IV. INFERENCE TO THE BEST LEGAL EXPLANATION

The idea of explanatory inference is the second main building block of the version of coherentism proposed.Footnote ³¹ Coherence may be construed in the course of legal decision making, I would argue, through an “inference to the best explanation.” That is, coherence-based legal inference is at bottom an explanatory kind of inference. Inference to the best explanation is the pattern of reasoning whereby explanatory hypotheses are generated and evaluated. More precisely, this inferential pattern may be defined as follows:

That is, an inference to the best explanation proceeds from a set of data to a hypothesis that explains the data better than any of the competing hypotheses. The mechanism we use to settle on which explanation to infer has three main stages: a process of generation, the result of which is a short list of plausible candidates; a context of pursuit, in which working hypotheses are subjected to a preliminary assessment and developed in further detail, and a process of selection, in which one of the competing hypotheses is accepted as justified.Footnote ³³ Coherence considerations are crucial in all three stages of explanatory inference.

First, coherence with background knowledge helps legal fact finders narrow down the range of plausible candidates. Explanations of the facts under dispute that fail to cohere with background beliefs, for example, beliefs about human motivations, causal principles, beliefs about how crimes are usually committed, and so on, are excluded from consideration. The process of generation, guided by this coherence constraint, results in the individuation of a “base of coherence,”Footnote ³⁴ that is, the set of hypotheses and evidence relevant to assessments of coherence, as well as the configuration of a “contrast set,”Footnote ³⁵ that is, the set of (initially plausible) alternative theories of the case.

Next, the pursuit of an initially plausible explanation of the facts under dispute involves rendering this hypothesis the best it can be. There are three main coherence-making strategies whereby promising lines of inquiry may be developed into full-blown hypotheses. First, the coherence of a theory of the case may be enhanced by “subtraction,” that is, by eliminating some beliefs. For example, faced with contradictory testimony, a fact finder may reach coherence by eliminating a belief in the credibility of one of the witnesses on the grounds that it conflicts with a hypothesis that is well supported by the available body of circumstantial evidence. Second, an “addition” strategy, whereby one or more beliefs are added, may also increase the coherence of a theory of the case.Footnote ³⁶ For instance, suppose that a legal decision maker believes that the evidence at trial strongly supports a guilt hypothesis. However, suppose that she also believes expert testimony that conflicts with the hypothesis of guilt. There emerge, however, in the course of the trial reasons for doubting the reliability of the method used by the expert. A fact finder may increase the coherence of the theory of the case entailing the guilt of the accused by adding the belief that the expert testimony is not reliable. Last, coherence may also be enhanced by a “reinterpretation” strategy, which amounts to eliminating one belief and replacing it by another.Footnote ³⁷ For instance, incriminating physical evidence found in the house of the accused can be reinterpreted, in light of evidence of police misconduct, as decreasing rather than enhancing the coherence of the theory of the case entailing guilt.Footnote ³⁸

Finally, coherence considerations are crucial in the context of justification. Among a set of robust theories of the case, fact finders are to select the one that is best, which, I would argue, should be understood as “best on a test of coherence.” In short, I suggest that the process of coherence maximization in the context of legal fact-finding is an explanatory inference that consists of the following steps:

1. i) The specification of a base of coherence, that is, the set of factual hypotheses and relevant evidence over which the coherence calculation proceeds;

2. ii) The construction of a contrast set that contains a number of alternative theories of the case from which the most coherent one is to be selected;

3. iii) The pursuit of the alternative theories of the case by means of a number of coherence-making mechanisms;

4. iv) The evaluation the coherence of the alternative theories of the case against the criteria of coherence that have been stated above;

5. v) The selection as justified of the most coherent of the alternative theories of the case, that is, the theory of the case that best satisfies the criteria of factual coherence.

The result of this process of coherence maximization is thus the selection as justified of the theory of the case that best coheres. Let me illustrate this process in terms of the example already introduced, that is, the Cannings case. The first stage is to specify the base of coherence. This involves specifying the set of hypotheses and evidence over which the coherence calculation is performed: for example, H₁ Angela Cannings killed her babies; H₂ Jason's and Matthew's deaths were natural; H₃ Jason and Matthew were killed by someone else; H₄ Angela Cannings was a loving mother; H₅ There is a genetic defect in the Cannings family, and so son.

At the second stage, the trier of fact specifies the set of theories of the case. Absent evidence that indicated the possibility that the babies might have been deliberately harmed by someone other than their mother, two theories of the case were considered. The first one, that is, “the Angela Cannings did it” theory, takes Lady Bracknell's approach as its starting point: “to lose one baby is a misfortune, two carelessness, three murder.”Footnote ³⁹ On this theory, the extreme rarity of three unexplained infant deaths in the same family grounds a powerful inference that deaths must have resulted from deliberate harm. The baby boys were smothered by someone who appeared to be an affectionate mother, and these deaths were related to a “pattern” of events in the family.

The second theory of the case, that is, “the natural deaths” theory, proceeds on the basis that the fact that there are deaths unexplained by our current state of knowledge does not exclude that there is a reasonable possibility that the deaths may be natural. Furthermore, it does not make sense to think that Angela Cannings had murdered her babies, for that would entail that she was becoming repeatedly pregnant with the sole purpose of killing the babies after they were born. Given that she was a loving mother and that there was no evidence that she had any personality disorder or psychiatric condition, the conclusion that she did kill her babies does not seem cogent.

The construction of the theories of the case involves determining the coherence and incoherence relations that hold between the different elements, that is, evidence and hypotheses. In order to do so, it is necessary to consider which evidence is explained by which hypothesis, whether there are any explanatory connections between competing hypotheses, whether there are relevant analogies at work, or whether there are any contradictions between propositions describing evidence or hypotheses. For example, the hypothesis that Angela Cannings did it coheres with expert evidence as to the rarity of three infant deaths in one family. And the hypothesis that there was a genetic defect in the Cannings family coheres with the hypothesis that the deaths were natural, as both are explanatorily related. By the end of this process, the fact finder would have constructed a number of theories of the case, each of which contains several hypotheses and evidence that cohere with each other, for example, the “Angela Cannings did it” and the “natural deaths” theories in this case.

At the third stage, the theories of the case under consideration are refined and revised with a view to making them the best that they can be. As argued above, there are three main strategies whereby coherence may be enhanced: additive, subtractive, and reinterpretive. For example, the fact that there was a third death in the family may be reinterpreted, in light of evidence of the possible existence of a genetic defect in the Cannings extended family, as evidence that lends support to the hypothesis that the babies’ deaths were natural rather than as evidence of guilt. As a result of this reinterpretation, the degree of coherence of the natural-deaths theory is enhanced. Or adding the belief that those who appear to be affectionate mothers sometimes kill their children increases the coherence of the Angela-Cannings-did-it theory by ironing out a possible inconsistency between the hypothesis that the babies were deliberately harmed and evidence that the babies were loved and cared for by their parents. This process of pursuit ensures that each competing theory of the case is developed and seriously considered before one takes any decision as to which of them is explanatorily best.

At the fourth stage, the fact finder assesses the coherence of the alternative theories of the case by examining the extent to which they satisfy the coherence constraints. She has to consider whether the guilt theory explains the evidence at trial or whether, to the contrary, the natural-deaths theory better explains the evidence available. She will also consider which of the theories is simpler and examine whether they fit with background knowledge about analogous cases. Sources of incoherence, that is, inconsistencies and incompatibilities, have to be identified as well. In the explanatory evaluation of the theories of the case, the fact finder has to give a priority both to the evidence admitted at trial as well as to the hypotheses that are compatible with innocence. The outcome of this evaluation is to arrive at comparative judgments of explanatory coherence. Although the theory according to which someone else did it lacks evidential support, the Angela Cannings-did-it theory and the natural-deaths theory both explain a substantial part of the evidence admitted at trial.

Finally, at the fifth stage, the most coherent theory of the case is selected, provided that its degree of justification (on this account, its degree of coherence) is high enough to satisfy the applicable standard of proof. In this case, I argue that as the Court of Appeals held, the acceptance of the Angela-did-it theory was not justified. First, there existed a coherent theory of innocence compatible with the evidence at trial. Most important, the body of expert evidence showed that although having three deaths in the same family was very rare indeed, there was a reasonable possibility that the deaths were natural. Second, some subset of the evidence is best explained by the innocence hypothesis, that is, evidence that Angela Cannings cared for and loved her children. Although that evidence is also consistent with a guilt hypothesis (it is possible to be both a loving mother and a baby killer), the innocence hypothesis coheres with it in a way that the guilt hypothesis does not. In light of these reasons, a coherence analysis would not sanction the acceptance of the guilt hypothesis. On appeal, this analysis was further fortified by the introduction of new evidence (expert evidence as well as evidence of other unexplained deaths in the Cannings extended family) that significantly increased the degree of coherence of the natural-deaths theory.

Now, this version of explanatory coherentism faces the following problem: there are cases in which the theory of the case that best satisfies the criteria of coherence is nonetheless intuitively unjustified. In other words, doubts may be raised as to whether satisfaction of the criteria of coherence is sufficient for justification. Two kinds of problem cases may be distinguished:

V. COHERENCE AND EPISTEMIC RESPONSIBILITY

I have argued above that the basic mechanism whereby judgments of coherence are made involves the generation of several alternatives and the selection among them of the one that coheres best. On this view, how well legal fact finders behave when generating and evaluating alternatives is relevant to determining how much reason they have for accepting as justified a theory of the case on the grounds that it is the most coherent one available. Thus, whether a fact finder is justified in accepting a hypothesis about the events being litigated is a matter that, at least in part, depends on whether she has considered all the relevant alternatives and assessed their coherence in a proper way.

This coherentist picture places the trier of fact, who seeks out coherence, at the center of a theory of justification. Coherence (and therefore justification) does not come for free, it is something that has to be earned. A coherentist legal epistemology turns out to be inextricably linked with a “responsibilist” view of justification, that is, the view that a belief is justified if it might be the product of an epistemically responsible action.Footnote ⁵⁴ The notion of “optimal coherence” is meant to capture and further develop this responsibilist dimension of coherentist justification. Optimal coherence—as defined before—is the kind of coherence that results from epistemically responsible coherence-based reasoning. An account is now due of what epistemic responsibility requires in the context of legal fact-finding.

We may distinguish between two main accounts of epistemic responsibility: a “deontic” approach and an “aretaic” approach. On a deontic approach, epistemic responsibility is a matter of duty fulfillment. That is to say, one is epistemically responsible to the extent that one complies with one's epistemological duties.Footnote ⁵⁵ According to the aretaic conception of epistemic responsibility, one is epistemically responsible insofar as one properly exercises a number of intellectual virtues.Footnote ⁵⁶ Epistemically responsible action amounts, on this view, to intellectually virtuous behavior.Footnote ⁵⁷ I endorse an irenic approach to the epistemic responsibility of legal fact finders, one that combines deontic and aretaic elements. The suggestion is that legal fact finders are epistemically responsible to the extent that they both comply with certain epistemological duties and exercise a number of intellectual virtues.Footnote ⁵⁸ Below I give an account of the epistemological duties whose satisfaction is required for the justification of evidentiary statements in law and of the intellectual virtues that mark off good epistemic behavior in the context of legal fact-finding.Footnote ⁵⁹

There are several epistemic duties compliance with which is necessary in order to conduct legal fact-finding in an epistemically responsible way. Legal fact finders have a duty to believe all and only the propositions that are supported by the evidence available.Footnote ⁶⁰ Thus triers of fact have a duty to withhold belief from what is not supported by the evidence at trial as well as a duty to believe when they do have good reason for so believing. In addition to the duty to believe as the evidence at trial dictates, fact finders also have a duty to gather additional evidence about propositions that are less than certain on the evidence available.Footnote ⁶¹ That is to say, fact finders are expected to seek out the evidence necessary to find out about the events at trial.Footnote ⁶² Triers of fact also have a duty to search actively for alternative hypotheses that have the potential to explain the events being litigated. In short, fact finders have a duty to behave in ways that will increase their chances of getting at the truth about the facts under dispute.Footnote ⁶³

The proper exercise of a number of intellectual virtues is also needed to live up to the standards of epistemic responsibility that govern legal fact-finding. Responsible legal fact-finding requires the exercise of the virtues of the good inquirer, such as thoroughness, sensitivity to detail, the ability to recognize salient facts, perseverance in following a line of investigation, diligence, the courage to entertain hypotheses that call in question deeply held beliefs, the ability to recognize reliable authorities, and open-mindedness in collecting and evaluating evidence. Fact finders should also exhibit a number of epistemic virtues when deliberating about the facts of the case. They ought to be impartial, that is, they are to assess the different alternatives on their merits and seriously consider the possibility that each might obtain. The virtues of open-mindedness and self-criticism are also critical here. Triers of fact ought to be open to new alternatives and ready to consider objections to their own views as well as to change their views in the face of new evidence or good arguments that might emerge in the course of deliberation. An epistemically responsible fact finder also displays the virtue of intellectual sobriety when deliberating about the case: she does not rush into judgment but engages in the evaluation of the competing hypotheses in a temperate and careful manner. Intellectual courage to face and answer criticism is necessary for virtuous deliberation as well. Finally, intellectual autonomy, that is, a capacity to develop one's own views, and intellectual humility, which involves the recognition of one's fallibility, are also necessary for reasoning about the facts under dispute in an epistemically responsible way.

To sum up, I have argued that the justification of conclusions about disputed questions of fact is a matter of optimal coherence. That is, a belief about the facts under dispute is justified if it belongs to the theory of the case that an epistemically responsible fact finder might hold by virtue of its coherence in like circumstances. Now, on this approach to the justification of evidentiary judgments in law, justification is the product of the effort of the fact finder who strives after coherence in an epistemically responsible manner. But, one may ask, how much work does one need to do in order to reach justified beliefs? What counts as a “reasonable” range of alternatives in the legal setting? How much evidence-seeking is necessary for epistemically responsible legal fact-finding? When is the final outcome coherent enough for the purposes of legal justification? Without an answer to these questions the proposed account of the justification of evidentiary conclusions in law remains hopelessly vague.

The suggestion that I want to pursue is that due attention to context gives us an answer to the foregoing questions. What qualifies as epistemically responsible behavior, and therefore the severity of the standards of justification, depends in an important sense, I shall argue, on a number of contextual factors. The claim is thus that contextual factors determine the appropriate level of care, that is, the appropriate level of epistemic responsibility that is required for legal justification. In other words, a coherentist approach to the justification of factual conclusions in law needs to be contextualized. The next section suggests some ways in which this might be done.

VI. COHERENCE IN CONTEXT

So far I have considered three elements of the coherence theory of evidence and proof: a concept of coherence, an explanationist account of coherence-based inference, and the idea of epistemic responsibility. A fourth ingredient is the recognition of the relevance of context to a coherentist epistemology. Assessments of coherence (and incoherence) are highly sensitive to context. Joseph Margolis writes:

Thus a critical part of the process whereby coherence is built consists in seeking out a relevant context in which assessments of coherence may be made. I suggest that we account for the context-sensitivity of judgments of coherence by “contextualizing” coherentism. Contextualism, that is, the view that standards of epistemic justification vary with context, is neutral as to the analysis of justification.Footnote ⁶⁵ In other words, it is not a theory about the nature of justification; rather, the contextualist tenet is that the severity of the standards of justification, however they might be analyzed, shifts with context. Thus there is room for a contextualized version of different accounts of justification. The suggestion is that coherentism should be contextualized. A contextualized version of coherentism would take it that whether a belief is justified is a matter of coherence but that how severe this coherentist standard should be is a contextually variable matter.Footnote ⁶⁶

Now, more precisely, what is it that shifts with context according to a contextualized version of coherentism? There are three dimensions upon which the severity of the coherentist standards of justification depends: i) the threshold of justification, that is, how strongly beliefs must cohere with one another to be justified; ii) the conditions of membership to the contrast set, that is, how many alternatives are relevant and should be ruled out for a belief in one of them to be justified; and iii) the domain of coherence, that is, how broad the domain should be within which coherence ought to be sought. The higher the threshold, the larger the contrast set; and the broader the domain of coherence, the more severe the standards of justification are.Footnote ⁶⁷ The main tenet of the contextualist approach to the coherentist standards for the justification of factual conclusions in law proposed here is that whether a theory of the case is coherent enough to be justified, how many alternative explanations of the facts under dispute ought to be ruled out, and what the appropriate domain of coherence is (i.e., how many hypotheses and how much evidence ought to be factored in the coherence calculation) shift with context. Contextual features help set the severity of the coherentist standards of legal justification by fixing the appropriate threshold, picking the relevant contrast set, and individuating the base or domain of coherence.Footnote ⁶⁸

Let us be more specific about which features of the context are relevant for the justification of evidentiary statements in law and how they fix the severity of the coherentist standards of legal justification. First, the stakes of the decision play a chief role in determining how severe the standards of justification ought to be. When the costs of being wrong are very high, a strict standard of justification is in order.Footnote ⁶⁹ The relevance that the importance of being right about a certain matter has to attributions of justified belief is recognized in law through the standards of proof. The threshold of justification of decisions in civil proceedings is typically lower than what is required for justification in a criminal proceeding. Furthermore, the very same standards of proof are context-dependent. For instance, what counts as reasonable doubt varies with the seriousness of the criminal offense.Footnote ⁷⁰ On a coherentist interpretation of these standards, this means that how much coherence is required for the justification of factual claims in law is context-dependent. Although a particular explanation of the facts under dispute may be coherent enough to justify a finding for the plaintiff in a civil case, it may fall below the threshold of coherence required to find against the defendant in a criminal case. And how many alternatives ought to be ruled out for a belief in a guilt hypothesis to be justified beyond a reasonable doubt by virtue of its coherence also depends on the gravity of the criminal charge. In cases involving serious criminal offenses, few explanations may be properly ignored for an inference to the most coherent of them to yield justified beliefs. The stakes are thus crucial for determining how demanding the coherentist standards of justification should be in legal contexts.Footnote ⁷¹

Second, the role that the legal decision maker occupies also determines the “level of scrutiny,” as Robert Fogelin puts it, which is appropriate in a particular context of justification.Footnote ⁷² Fact-finding in law involves different legal actors, for example, police, prosecutors, trial judges, jurors, and appellate judges. A higher level of scrutiny is appropriate for different roles. Although an explanation of the facts under dispute that entails guilt might be coherent enough for the purposes of filing a criminal charge, it might not be so in the context of jury decision making. How much evidence and how many alternative hypotheses need to be factored into a coherence calculation also depends on the institutional role the fact finder occupies. The domain within which coherence ought to be sought increases the more the decision task inherent to a particular role is likely to affect the parties involved (in ways that cannot be remedied later in the process).

Third, various goals are relevant to different contexts of justification and relative to which a belief may be properly characterized as justified.Footnote ⁷³ In law, a complex set of goals, which the rules of evidence are meant to protect, have a bearing on questions of justification. The goals of inquiry help determine the appropriate domain of coherence. For example, prejudicial evidence ought to be excluded from the base of coherence in legal contexts, whereas it would be appropriate and indeed required to include it in contexts in which, unlike the legal one, the only goal of inquiry is the acquisition of true beliefs. To illustrate the point, the mass of evidence that is relevant to historical research on a controversial murder case (e.g., the famous Sacco-Vanzetti case) is different from the evidence on the basis of which a belief about the guilt of the defendant may be justified, given that values other than truth are at stake in the legal context.

Fourth, different methodological constraints, which determine the direction of inquiry, are in place in different contexts.Footnote ⁷⁴ What is at stake here, as Michael Williams puts it, is not so much the “level” of scrutiny as the “angle” of scrutiny. He writes, “Within the practice of doing history, we can be more or less strict in our standards of evidence. But some questions have to be set aside for us to think historically at all.”Footnote ⁷⁵ Similarly, the fact that a question of legal justification is being raised constrains the range of alternatives that ought to be seriously considered. The following example by Walter Sinnott-Armstrong vividly illustrates the point: “Suppose that a desperate defense attorney argues, ‘Perhaps we are all deceived by a demon.’ Surely the judge and jury could attend to this statement and yet properly ignore it by not letting it affect their belief in the defendant's guilt.”Footnote ⁷⁶ Indeed, although the possibility that we are all deceived by a demon is a relevant possibility in the context of epistemological inquiry, it can be properly ignored in the context of legal inquiry.

Methodological constraints also contribute to determining the domain of coherence by requiring that some elements be included and apportioned a special weight in the coherence calculation. For example, evidence at trial has to be taken into account and given a special priority for one to be engaged in legal reasoning. Similarly, the irrelevance of other considerations (e.g., character evidence) must also be taken for granted for one to argue from a legal point of view. Failure to comply with these methodological constraints when performing coherence-based reasoning about facts in legal contexts would amount to changing the nature of inquiry altogether.

Fifth, the level of scrutiny that is reasonable in a particular context depends on the resources available. Legal fact-finding works under severe time constraints and has limited institutional resources at its disposal. These constraints limit the domain within which coherence ought to be sought and the range of alternatives that need to be considered when reasoning about evidence in law, as opposed to the much broader inquiry that would be appropriate in other contexts, like, for example, the scientific context.

Last, justificatory practices take place in a dialectical context that constrains what may be taken for granted and what, to the contrary, is a relevant alternative that needs to be ruled out for one's claim to be justified.Footnote ⁷⁷ The fact of raising a possible defeater triggers a higher level of scrutiny. For example, an expert's testimony may not be taken at face value as soon as doubts are raised about the credibility of the expert. Simply mentioning a possible defeater to one's claim, no matter how remote, is not, however, enough to raise the standards of justification.Footnote ⁷⁸ Only those alternatives of which we have some reason to believe that they obtain or at least some positive reason to doubt that they are false are relevant, and should thus be attended, in a particular context.Footnote ⁷⁹ For example, consider a witness, John, who testifies to the effect that he saw the defendant committing the crime. The possibility that John was hallucinating that the defendant committed the crime is a possibility that the fact finder may properly exclude from the contrast set. However, once the defense attorney claims that at the moment of the crime John was under the effects of a drug that may cause hallucinations as a side effect, this possibility becomes a relevant possibility—as long as there is some positive reason to believe that it might obtain—that needs to be included within the contrast set. Thus dialectical constraints shape the contrast set by making some alternatives salient.

The foregoing contextual features, among others, determine the severity of the coherentist standards of legal justification. In so doing, they establish what is required for responsible action when reasoning about evidence in law. In other words, they set a “threshold of termination” of the process of coherence maximization by determining when it is legitimate to stop the quest for coherence.Footnote ⁸⁰ Once fact finders have reached that point by building coherence among the hypotheses and evidence that are relevant in the particular context of justification, they may be said to have done all that can be expected of them to do, that is, all that is required for justification.

To recapitulate, I have argued that a legal decision maker's belief about the facts under dispute is justified if it could result from an epistemically responsible process of coherence maximization. Coherence is a matter of the satisfaction of a number of positive and negative constraints among the relevant hypotheses and evidence at trial. Such coherence is constructed in the course of legal decision making by means of an inference to the best explanation. This explanatory inference yields justified beliefs provided that some standards of epistemic responsibility are complied with. Just what these standards require is a matter that depends, I have claimed, on context. In short, this coherentist approach to the standards of justification of evidentiary judgments in law is based on a constraint-satisfaction conception of coherence, an explanatory view of coherence-based legal inference, and a responsibilist approach to justification that recognizes the relevance of context to judgments of coherence and justification. In the next section, I examine how the proposed coherentist account of the justification of conclusions about disputed questions of fact deals with a number of objections that threaten to undermine the case for a coherence theory of evidence and juridical proof.

VII. SOME OBJECTIONS TO A COHERENCE THEORY OF EVIDENCE AND LEGAL PROOF

A number of objections may be raised against attempts to analyze the justification of evidentiary judgments in law in terms of coherence. At bottom, these problems are but particular instances of the main objections that have been traditionally raised against coherence theories of epistemic justification. A satisfactory response to these problems is necessary for the coherence theory to be a plausible candidate for a theory of the justification of evidentiary judgments in law. In what follows, I suggest some ways in which the proposed theory of legal justification as optimal coherence may meet the objections that may be leveled against coherence theories of evidence and proof in law.

VIII. TOWARD A JURISPRUDENCE OF EVIDENCE

In the sections above I have articulated and defended a coherence theory of the justification of evidentiary judgments in law, that is, the theory of optimal coherence. This coherentist approach puts forward a picture of legal epistemology that has the following prominent features. First, the theory of optimal coherence is of a piece with a naturalistic approach to issues of evidence and proof in law, which takes facts about how fact finders reason to be relevant to answering questions about how they ought to reason.Footnote ¹⁰⁰ From a naturalistic standpoint, coherentism is initially attractive, as our cognitive endowment seems to be geared toward coherence.Footnote ¹⁰¹ There is substantial evidence establishing that coherence plays a significant role in legal fact-finding.Footnote ¹⁰² Moreover, constraint satisfaction mechanisms have been shown to be at work in reasoning about evidence in conditions of complexity, such as those that characterize the context of legal fact-finding.Footnote ¹⁰³ Hence the version of legal coherentism proposed advances the naturalistic project of devising theories of evidential reasoning that can be implemented by real rather than ideal legal agents.

Second, the theory of optimal coherence endorses a contextualist approach to legal epistemology according to which the justification of conclusions about disputed questions of fact is context-dependent. This trend toward contextualizing coherentism goes hand in hand with the development of a naturalized epistemology of legal proof. A contextualist understanding of the set of beliefs that confers justification significantly increases the psychological plausibility of the coherence theory of justification, as it does not require legal fact finders to build coherence among the whole system of beliefs, which is beyond what they may achieve given our cognitive and memory resources, but, more modestly, it expects fact finders to achieve coherence only within the subset of beliefs that is shown to be relevant in the particular context.

Third, the theory of optimal coherence puts forward a responsibilist conception of the justification of evidentiary judgments in law, which claims that the standards of behavior that regulate inquiry and deliberation about factual matters in law have a bearing on the justificatory status of the resulting decisions. More specifically, a main tenet of the model of coherence proposed here is that compliance with some epistemic duties and the exercise of some epistemic virtues are necessary for ascribing justified beliefs in the legal context. It follows from this view that legal epistemology is best understood within the broader framework of “virtue epistemology,” which makes virtues central to the analysis of justification and knowledge.Footnote ¹⁰⁴ From this perspective, the study of the features that mark off virtuous behavior in legal fact-finding is not merely the appropriate subject of legal ethics but rather a substantial part of a jurisprudence of evidence.

Fourth, the coherentist model developed in this paper underwrites an explanationist approach to evidential reasoning in law, according to which legal reasoning about evidence is first and foremost an explanatory kind of reasoning. An important consequence of turning to explanationism is that of expanding the scope of the theory of legal reasoning beyond its traditional boundaries. Questions pertaining to the context of discovery are largely neglected by theories of legal reasoning, whose focus is on questions of justification. In the context of evidence law, this reductive view has been reinforced, for the dominant probabilistic paradigm—that is, Bayesianism—is not applicable to discovery-related issues.Footnote ¹⁰⁵ In contrast, explanatory inference is a kind of inference whereby hypotheses are both generated and evaluated. Thus an explanatory approach to factual inference in law places both questions of discovery and questions of justification within the scope of a theory of evidential reasoning in law. As a result, it expands the role of reason in law by making discovery, as much as evaluation, a process that falls within the realm of reason.

Last, the coherence model articulated above advances a noninstrumentalist approach to reasoning about evidence in law, according to which deliberation about factual matters involves deliberation about ends. This is a second way in which the coherentist approach to justification moves toward expanding the scope of reason in factual decision making in law. Fact-finding is part of a broader process of decision making that is meant to achieve the satisfaction of different and often conflicting ends. But if this is so, then a theory of evidential reasoning in law should not merely tell us—as is implicitly assumed—what are the best means to achieve some fixed ends. Rather, it should also provide legal decision makers with guidance as to how to proceed in the face of value conflict and, most important, how to strike a balance between truth and other values. The coherentist approach defended here satisfies this demand by placing questions about the ends of adjudication within the reach of deliberation.

To conclude, in this paper I have argued for a coherentist approach to the justification of evidentiary judgments in law according to which justification is a matter of optimal coherence. I have shown that this theory overcomes some of the main problems facing coherence theories of evidence and proof. In addition, it has some interesting implications for the theory of evidential reasoning in law. Nonetheless, there are still many aspects of the proposed version of legal coherentism that need to be further developed.

First, a constraint-satisfaction characterization of factual coherence faces the “problem of the input,” that is, the problem of determining the elements and the coherence (and incoherence) relations upon which assessments of coherence depend. In addition, it also seems necessary to clarify further how the different kinds of coherence that are relevant to the justification of conclusions about disputed questions of fact relate to each other, that is, the “problem of integration.”

Second, the theory of optimal coherence rests on a naturalized view of legal epistemology. The objection that naturalized approaches to epistemology fail to account for the normativity of justification applies equally to any attempt to naturalize the epistemology of legal proof. Thus it is necessary to spell out in detail the kind of naturalism involved in this approach to coherentism so as to put to rest worries about whether it preserves the normative dimension of legal justification.

Third, some versions of contextualism lead to forms of relativism or skepticism that are not acceptable in the context of legal fact-finding. The question arises as to whether contextualizing coherentism in the way proposed here leads to an undesirable relativization of the conditions under which evidentiary judgments in law are justified.

Fourth, the coherentist theory of evidence and proof developed in this paper makes virtues and duties central to issues of justification. This responsibilist approach to legal epistemology is in its infancy. A more thorough account of how epistemically responsible behavior relates to the justification of factual conclusions in law is surely needed. More work also needs to be done with regard to the kind of epistemic virtues and duties that are relevant to legal fact-finding as well as how they relate to moral virtues and duties.

Last, much more needs to be said about how explanatory inference works in the legal setting. Despite all the work that still remains to be done, I hope that this paper contributes to advancing the study of the justification of evidentiary conclusions in law and to establishing coherentism as a plausible candidate worth exploring further.