Legal theorists in the U.S. do devote much attention uncertainty in the law, but “theoretical” law teachers in the U.S. worry far more about legal uncertainty than about factual uncertainty. In American legal literature in the last quarter of the 20^th century, the phenomenon or problem legal uncertainty has often been referred to as the phenomenon or problem of “legal indeterminacy.” This linguistic usage came into vogue as the “Critical Legal Studies” movement came into vogue. The literature on “legal indeterminacy” is now vast. See, e.g., the following very small sample (largely borrowed from the article by G. Lawson cited further below) of the literature on “legal indeterminacy”: Anthony D'Amato, Can Any Legal Theory Constrain Any Judicial Decision?, 43 University of Miami Law Review 513 (1989); Anthony D'Amato, Can Legislatures Constrain Judicial Interpretation of Statutes?, 75 Virginia Law Review 561 (1989); Anthony D'Amato, Pragmatic Indeterminacy, 85 Northwestern University Law Review 148 (1990); Stanley Fish, Dennis Martinez and the Uses of Theory, 96 Yale Law Journal 1773 (1987); Stanley Fish, Don't Know Much About the Middle Ages: Posner on Law and Literature, 97 Yale L.J. 777 (1988); Ken Kress, A Preface to Epistemological Indeterminacy, 85 Northwestern University Law Review 134 (1990); Ken Kress, Legal Indeterminacy, 77 California Law Review 283, 286 (1989); J. Lipkin, Beyond Skepticism, Foundationalism and the New Fuzziness: The Role of Wide Reflective Equilibrium in Legal Theory, 75 Cornell Law Review 811 (1990); Frederick Schauer, Easy Cases, 58 S. California Law Review 399 (1985); Joseph Singer,The Player and the Cards: Nihilism and Legal Theory, 94 Yale L.J. 1 (1984); Steven L. Winter, Bull Durham and the Uses of Theory, 42 Stanford Law Review 639 (1990).

When American legal theorists speak of “legal indeterminacy,” they are referring, in the main, to the problem of the uncertain meaning of legal rules, they are alluding to semantic uncertainty in legal rules, they are directing attention to uncertainty of propositions about the meaning of legal rules. When legal theorists refer to “legal indeterminacy,” they generally are not trying to direct attention to the possible difficulty of determining whether the facts or circumstances that warrant or license the application of a legal rule are or are not present.

Interpretation – the use of particular techniques for determining1the meaning of legal pronouncements in source materials such as statute books and judicial opinions – is a prominent topic in legal scholarship. The reason for this is at least in part due to the importance that society as a whole attaches to keeping levels of uncertainty about the existence and the meaning of legal rules at some “tolerably” low level.

There have been scattered exceptions to the pervasive indifference of U.S. legal theorists about the pervasive phenomenon of factual uncertainty. See, e.g., Jerome Frank, Law and the Modern Mind (1^st ed., 1930). See also J. Frank, Courts on Trial: Myth and Reality in American Justice (1949).

During the last 30 years an important movement known as the “new evidence scholarship” has emerged. Some of the participants in this movement or tendency engage in what can only be called “philosophical” analysis of factual uncertainty and inference. However, until recently, this movement has had relatively little impact on the sort of “general” legal theorizing that has been done by scholars such Ronald Dworkin and H.L.A. Hart. Please see Section 2 of this Handout for references to literature in and about the “new evidence scholarship.”

In my Konstanz lectures I will not try to discuss of all forms of uncertainty in law. (It would be stupid and vainglorious of me to tackle so broad a topic!)  I will focus instead on uncertain inference about factual hypotheses. Hence, I will skirt the topic of inconclusive reasoning about legal rules.

My decision to focus on factual uncertainty in law might be challenged on the ground that my decision rests on an invalid presupposition: the presupposition that there is a difference between factual hypotheses and legal hypotheses.

The distinction between law and fact is common in U.S. law. The distinction is often also frequently made by U.S. law teachers. To be sure, it is frequently acknowledged both in U.S. legal doctrine and by U.S. legal theorists that questions of fact and questions of law are not entirely disjoint. Nonetheless, it is generally accepted in U.S. legal circles that questions of fact and questions of law are at least substantially disjoint .

Despite the nearly-unanimous view that there is some meat to the law-fact distinction, there is one important respect in which it is far from certain that the conventional disjunction between factual hypotheses and legal hypotheses – the conventional disjunction, we might say, between hypotheses about spatio-temporal events and hypotheses about legal rules and doctrines –, there is one important respect in which it is uncertain that the putative law-fact distinction exists:

It is far from certain that  the set of propositions about the world and set of propositions about legal rules are disjoint – because it is far from certain that propositions about legal rules can be said to be propositions that do not belong to the class of propositions that are about the world.

Given the dubious epistemological and ontological foundations of the law-fact distinction, it is not surprising that the customary distinction between law and fact has not always been entirely uncontroversial, and it is not surprising that some controversy about the alleged distinction continues to this day.

In recent years Mirjan Damaška has argued convincingly that both in medieval English law and in medieval occidental law in general the distinction between law and fact both was either non-existent or very diffuse. See, e.g., M. Damaška, Rational and Irrational Proof Revisited, in J.F. Nijboer & J.M. Reijntjes, eds., Proceedings of the First World Conference on New Trends in Criminal Investigation and Evidence 75 (1997) (proceedings of 1995 conference).

There are strands in Hans Kelsen’s philosophy of law and in the school of legal philosophy known as “Scandinavian Realism” that suggest that at least some legal theorists who swim in the general intellectual tradition known as “legal positivism” believe (wittingly or unwittingly) that even if legal norms belong to a category of existence or knowledge that is somehow separate from the realm of “facts,” methods of determining the existence and meaning of norms and facts are or ought to be similar. See generally, e.g., Hans Kelsen, General Theory of Law and State (1945); Karl Olivecrona, Law as Fact (1^st ed., 1939; 2^nd  ed., 1971).

A few legal theorists in the United States have frontally challenged the law-fact distinction. See, e.g., Peter Tillers, The Value of Evidence in Law, 39 Northern Ireland Law Quarterly 167 (1988);Gary Lawson, Proving Law, 86 Northwestern University Law Review 859, 863 (1992); John Duffy, On Improving the Legal Process of Claim Interpretation: Administrative Alternatives, 2 Washington Journal of Law 109, 122 (2000).

Several distinct types of challenges might be mounted agaimst the law-fact distinction. The most modest challenge would rest on the thesis that the set of propositions about legal rules and the set of propositions about spatio-temporal events overlap – and that, thus, some propositions about legal rules are also propositions about spatio-temporal events. A second and bolder challenge to the law-fact distinction would rest on the thesis that one or both of these sets of propositions – the set of propositions about law and the set of proposition about spatio-temporal states – not only overlap, but are also very fuzzy or rough – “indeterminate” to a substantial degree. (The critique of M. Damaška, mentioned above, seems to fall roughly into this middle category.) The most  radical challenge to the distinction would presumably rely on the thesis that all propositions about legal rules are propositions about spatio-temporal states of affairs and that there is no fundamental difference between propositions about legal rules in the spatio-temporal world and other states of the spatio-temporal framework. (An alternative form of this radical thesis might be that the two types of propositions are completely fuzzy or rough – that there is no there there.)

I readily acknowledge that both the nature of law-fact distinction and its existence are uncertain. For purposes of my Konstanz lectures, however, I posit that the law-fact distinction holds water. I can do so without harming my argument because even if the distinction fails – even if the disjunction between the two kinds of hypotheses completely collapses! – my analysis of factual inference remains valid and useful: the web-like structure of evidence and inference that I describe in my lectures remains intact – it continues to exist – even if law is fact and even if argument about law is much like argument about fact. (Indeed, if it turns out that law is fact, and that reasoning about law is (essentially) the same as reasoning about fact, the “bottom line” of my entire argument – which to a large extent rests on the hypothesis of the deleterious effects of time on the strength and stability of uncertain inference in law –, to that extent the bottom line of my argument gains force.)

There is reason to think that I was invited to take part in the Konstanz summer school largely because of my role in a movement or intellectual tendency that is widely known as the “new evidence scholarship.” (Otherwise the existence of the gracious the invitation that I received is utterly inexplicable!)

The “new evidence scholarship” – NES – is an academic movement whose origins lie principally in the U.S and in the U.K. (Some U.K. people think the origins of NES lie more in the U.K. than in the U.S. I think the reverse is the case. It hardly matters who is right about this question. [But some observers think that the origins of the new evidence scholarship must be traced to antiquity. Here I must draw the line: the effect of this genetic approach would turn the label “new evidence scholarship” into an oxymoron!])

NES is a broad academic movement or tendency that has no single defining characteristic. Moreover, it would be very difficult even to lay out a set of characteristics that define this movement or tendency. A few things, however, can be said with confidence about the movement:

NES focuses on factual inference and proof in legal settings.

Traditional U.S. legal scholarship about the law of evidence focuses on legal rules governing the admissibility of different kinds of evidence in trials, but NES focuses on such rules much less and devotes much more attention to procedures for assessing the weight of evidence admitted in trials and other legal proceedings.

NES devotes a considerable amount of attention to the logic of inference about factual hypothesis.

A significant chunk of NES literature makes use of formal probability theory and statistical methods.

One way to think of the new evidence scholarship is to see it as a movement that began largely in response to a renowned article by a Harvard Law School professor, Laurence H. Tribe: Trial by Mathematics: Precision and Ritual in the Legal Process, 84 Harvard Law Review 1329 (1971). Tribe’s article attacked a proposal to use Bayes’ Theorem to resolve certain kinds of factual questions in trials.

The moniker “new evidence scholarship” was coined by Richard Lempert. See Richard Lempert, The New Evidence Scholarship: Analyzing the Process of Proof, 66 Boston University Law Review 439 (1986). There is now a vast body of literature that might fairly be characterized as either being “new evidence scholarship” or as being about this new kind of scholarship. See, e.g., Symposium, Decision and Inference in Litigation, 13 Cardozo Law Review 253 (1991); Symposium, Probability and Inference in the Law of Evidence, 66 Boston University Law Review 377 (1986); William Twining, Rethinking Evidence (1990); William Twining, Theories of Evidence: Bentham and Wigmore (1985); David Schum, Evidence and Inference for the Intelligence Analyst (1987) (two volumes);  Richard D. Friedman, Route Analysis of Credibility and Hearsay, 96 Yale Law Journal 667 (1987); Richard Friedman, Character Impeachment Evidence: Psycho-Bayesian [!?] Analysis and a Proposed Overhaul, 38 UCLA L. Rev. 637 (1991); Roger C. Park, Evidence Scholarship, Old and New, 75 Minnesota Law Review 849 (1991); William Twining, The New Evidence Scholarship, 13 Cardozo Law Rev. 295 (1991); Daniel Shaviro, Statistical-Probability Evidence and the Appearance of Justice, 103 Harvard Law Review 530 (1989); Bernard Robertson & G.A. Vignaux, Interpreting Evidence: Evaluating Forensic Science in the Courtroom (1995); Joseph B. Kadane & David A. Schum, A Probabilistic Analysis of the Sacco and Vanzetti Evidence (1996); David Schum, Evidential Foundations of Probabilistic Reasoning (1994); J.F. Nijboer & J.M. Reijntjes, eds., Proceedings of the First World Conference on New Trends in Criminal Investigation and Evidence 75 (1997); Richard D. Friedman, Answering the Bayesioskeptical Challenge, 1 Int’l Journal of Evidence & Proof 276 (1997); Dale A. Nance, Evidential Completeness and the Burden of Proof, 49 Hastings Law Journal 621 (1998); Richard A. Posner, An Economic Approach to the Law of Evidence, 51 Stanford Law Review 1477 (1999); Richard D. Friedman, A Presumption of Innocence, Not of Even Odds, 52 Stanford Law Review 873 (2000); Ronald J. Allen & Brian Leiter, Naturalized Epistemology and the Law of Evidence, 87 Virginia Law Review 1491 (2001); Dale A. Nance, Naturalized Epistemology and the Critique of Evidence Theory, 87 Virginia Law Review 1551 (2001).

It is possible that I am here in Konstanz under false pretenses because I will have little to say about many or most of the issues that have occupied the attention of “new evidence scholars.” But I will talk about uncertain factual inference and uncertain proof in legal settings. So perhaps there is some connection between the preoccupations of new evidence scholars and the matters that I will discuss.2 (But if you conclude that the connection is tenuous, you might well have reason and evidence on your side!)

In one of my Konstanz lectures I talk a bit about the question of why the graphs or diagrams that legal theorists and their admirers construct to depict inference tend to have directed arcs that run from evidence to (ultimate) hypotheses rather than vice-versa. I pose this question but in my lectures I deliberately avoid attacking this question directly. While I cannot answer my own question in this handout, I can point to the beginnings of several possible answers. The possible explanations that I can discern are as follows:

i. Legal proof even at trial remains to some extent abductive rather than deductive (or retroductive). Proof remains abductive – suggestive of hypotheses --  because even at a trial the question of which factual hypotheses are or ought to be at issue often remains in play, open -- because of, e.g., the possibility in the U.S. of amending pleadings in civil cases to conform to evidence and because of, e.g., the generality of the language frequently found in both civil and criminal pleadings}. Thus, it might be argued, a significant part of the function of evidence even at trial is to suggest hypotheses as well to substantiate known hypotheses. This abductive function of evidence – this suggestive function, this hypothesis-forming function – is better illustrated by arcs that run from evidence to factual hypotheses rather than vice-versa.

ii. At a trial – at least at a typical “concentrated” American trial – the possibility of gathering additional evidence is largely foreclosed. Hence, although it is possible that evidence at trial has a legitimate hypothesis-formation function (see par. no. i above), once a U.S.-style trial starts, the parties to a lawsuit ordinarily cannot gather much evidence in addition to the evidence that is already in their possession when the trial starts. This (it might be argued) is a good reason for the bottom-up perspective of legal professionals – because once a collection of evidence is closed, there is less reason than there otherwise would be to examine the ultimate propositions at issue for the purposing of deducing or inferring the additional evidence that might or should be found were additional investigations or observations taken.

Even if this explanation makes some sense, its reach is easily exaggerated: even with the existence of extensive pretrial [legal] discovery rights [in civil litigation] many unpredictable events happen at every trial and much unanticipated and unforeseen evidence is generated and submitted at every trial – because before a trial ends no party has perfect knowledge -- or even nearly-perfect knowledge -- of the evidence that will ultimately end up being submitted at any trial. (Keep this in mind: a wink of an eye is “evidence,” a frown is “evidence,” a turn of a phrase is “evidence,” and similar minor curlicues in evidence are frequently, and perhaps ordinarily, very significant bits of “new” evidence.)

iii. There is a possibility that legal professionals – and actors such as jurors – think about evidence in non-Bayesian ways. For example, perhaps L.J. Cohen is correct: perhaps they think (at least sometimes) in a “Baconian” fashion.

Counterpoint by question: But would the use of Baconian reasoning by decision makers in litigation really change anything? Granted, L.J. Cohen has argued that Baconian reasoning involves a quest for unconditional judgments about possible facts. But isn’t this notion just plainly and flatly wrong? Even if you think that it is appropriate or possible to reason about evidence and  factual hypotheses in a Baconian or ordinal way, is it not the case that even a Baconian must be interested in the judgment that should be made about some question of fact given a particular collection of evidence, not given nothing? If so, is it not the case that even a Baconian reasoner faces a problem of conditional inference I(H|E) – the inductive strength I of hypothesis H given evidence E, albeit  not a  judgment about a conditional probability? [So it seems to me.]

iv. Perhaps there is a stronger tendency and also more reason to have arcs run ”upwards” from evidence to hypothesis when or to the extent that it becomes difficult to deduce or retroduce the available or existing evidence from the hypotheses in play. The notion here is that there is no good reason to use a top-down deductive or retroductive mode of thought when it is the case that a deductive or retroductive pattern of thought or reasoning would generate a wide range of expected evidence or observations but only a limited amount (subset) of that potential evidence is in hand and it is difficult or impossible to gather more.

Qualification by question: Isn’t this explanation just a variant of the explanation found in par. no. i above?

v. In some contexts, there is value in making models or pictures of inference that serve psycho-logical as well as logical purposes. If the direction of arcs in an inference network is a matter of  indifference from a logical or mathematical perspective, in some contexts it makes sense to make the arcs run in the direction that comports best with psychology, common conceptions. When all else is equal, make inference networks user-friendly! And – moreover – possibly, just possibly, there is sometimes good reason to make inference networks user-friendly even if one mode of representation is more congenial to computational logic than is some other  (more user-friendly) method of representation: perhaps, in short, sometimes the “rational” thing to do is to sacrifice some of the interests of logic – particularly the interests of computational logic – particularly when computation, for one reason or another, is beyond the ken or yen of the reasoner. Perhaps, in short, it sometimes makes sense to say that sometimes we make logic stronger by making it weaker.

I am making a radical suggestion. The sorts of inference networks to which I am now alluding make computation and calculation quite difficult and, perhaps for that reason (or perhaps for another reason), few parts of the kinds of inference networks I sketch can safely be trusted to run on their own – they seem to require a guiding human hand at almost every point – and so perhaps, for that reason, the old charge that this softer or mushier form of reasoning, the kind to which I allude (and which I describe a bit in my lectures), is not really reasoning at all is a valid charge – because it may in fact be very hard to point to instances in which the suggested mushy mode of argument to which I allude forces anyone’s mental hand.  But perhaps it is still true – this is the BIG question – that the mushy form of inferential reasoning to which I allude (and that I describe in my lectures) still somehow, almost mysteriously, constrains or guides thought and, thus, possibly, leads to better judgments and decisions. But I cannot swear that this is true: I do not know enough to venture a final judgment on this BIG question! What say you, dear Reader? (Or does it seem to you that my general question here is not even intelligible, that my question is incoherent?)

N.B. David Schum offers the following explanations for his strong preference for graphs with arcs that run upwards, from evidence to hypotheses:
 

[In a standard Bayes net] the arcs go … downward to the evidence. This is the result of an early convention that was established by persons designing Bayesian network software systems such as INDIA and ERGO. In part, this convention can be associated with the work of Judea Pearl, whose work on inference networks formed the basis for much of the later software developments. (D. Schum, Alternative Views of Argument Construction from a Mass of Evidence, in M. MacCrimmon & P. Tillers, eds., The Dynamics of Judicial Proof: Computation, Logic, and Common Sense 169 (2002))

Wigmore charts … certainly qualify as far as the DAG [directed acyclic graph] requirement [articulated by Judea Pearl] is concerned, but there are some differences between Wigmore charts and Bayesian networks as they are employed in process modeling. One obvious difference involves the direction of reasoning shown in the two forms of analysis. … Notice that Pearl mentions that the links among nodes in a Bayesian network indicate “informational or causal” dependencies. Here is the basic difference between the two. Wigmore charts are constructed to show the relevance of evidence on some stated ultimate and penultimate probanda. This accounts for the “upward” reasoning (from evidence to probanda) on a Wigmore chart. I have argued elsewhere that a relevance relation and a causal relation are not the same. (Id. at p. 173)

In short, Schum seems to think that the reason for downward arcs in Bayesian networks is (i) historical accident and (ii) the mistaken view that relevance relations always track causal relations.

1. Legal scholars often implicitly or explicitly welcome the ambiguity of the word “determine,” a word that, like the German word “bestimmen,” can have two quite distinct connotations or senses: (i)  uncovering or discovering an existing meaning, and (ii) endowing or conferring a meaning that previously – before Bestimmung – may not have existed.