DISCOVERING THE LOGIC

OF LEGAL REASONING
Vern R. Walker*

I.

INTRODUCTION

The rule of law rests on the quality of legal reasoning. The rule of
law requires that similar cases should be decided similarly, that each
case should be decided on its merits, and that decision-making processes
should comply with applicable rules of procedure and evidence. Making
the reasoning behind such decision-making transparent and open to
scrutiny shifts the decisions away from mere subjective preference and
toward objective rationale. An important means, therefore, of achieving
the rule of law is articulating and evaluating the various elements of
legal reasoningthe reasoning involved in interpreting constitutions,
statutes, and regulations, in balancing fundamental principles and
policies, in adopting and modifying legal rules, in applying those rules to
cases, in evaluating evidence, and in making ultimate decisions.
Despite our need for transparent and sound reasoning, we in the
legal profession devote surprisingly little research to developing our own
general methodology. This is in dramatic contrast to other fields and
professions. We are not like mathematicians, whose reflection on their
own method has given the world axiomatic proof and modern deductive
logic. We are also unlike statisticians, who have developed the analytic
methods in use in all areas of empirical research. Nor do we act like
natural and social scientists, who carefully combine statistical methods
with techniques for measurement and modeling that are tailored to their
particular subject matters. Nor do we take the approach of the medical
profession, which has refined various methodologies for diagnostic
reasoning. We in the legal profession largely content ourselves with
knowing good legal reasoning when we see it. We spend relatively
little time refining general methods for discriminating between good
patterns of reasoning and bad, or developing theories for explaining
precisely why good patterns are good and bad patterns are bad. In sum,
we do not pay particular attention to the logic of legal reasoning.1 For a
* Professor of Law, Hofstra University School of Law; Hempstead, New York, USA. The
author is grateful for the financial support provided by a research grant from Hofstra University.
1. The word logic refers here to the study of the methods and principles used to
distinguish correct reasoning from incorrect reasoning, and to the theories that result from such
study. See IRVING M. COPI & CARL COHEN, INTRODUCTION TO LOGIC 3 (10th ed. 1998). Correct
reasoning warrants the conclusion to be true, probably true, or at least plausible. It provides

1687

1688

HOFSTRA LAW REVIEW

[Vol. 35:1687

profession so dependent upon constructing good arguments, we are

oddly uninterested in good methodology for argument construction.
There certainly exist isolated pockets of theoretical work on legal
reasoning methods. Those involved in criminal law study the probative
value of the forensic sciences,2 and the Supreme Courts trilogy of cases
beginning with Daubert has spawned extensive research on when expert
opinions are good enough to be admissible.3 A few theorists since
Wigmore have also studied general patterns of evidence evaluation.4
Fields outside of law have also conducted important research on various
aspects of legal reasoningfields such as psychology,5 rhetoric,6
informal logic,7 and artificial intelligence (AI).8 What is missing,
adequate justification for a reasonable persons adoption of the conclusion. Logic is distinct from
the study of methods for discovering correct lines of reasoning (for example, heuristics), although
logic can help identify the desired goal or end product of heuristic methods. Logic is also distinct
from the study of persuasive use of reasoning in human dialogue (for example, pragmatics, rhetoric,
or psychology), although it can help identify a reasonable basis for persuasion. Logic is the study of
how we ought to reason, if our goal is to discover truth.
2. See, e.g., DAVID L. FAIGMAN, DAVID H. KAYE, MICHAEL J. SAKS & JOSEPH SANDERS,
SCIENCE IN THE LAW: FORENSIC SCIENCE ISSUES (2002).
3. Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579 (1993); Gen. Elec. Co. v. Joiner, 522
U.S. 136 (1997); Kumho Tire Co. v. Carmichael, 526 U.S. 137 (1999).
4. E.g., TERENCE ANDERSON, DAVID SCHUM & WILLIAM TWINING, ANALYSIS OF EVIDENCE
(2d ed. 2005); JOSEPH B. KADANE & DAVID A. SCHUM, A PROBABILISTIC ANALYSIS OF THE SACCO
AND VANZETTI EVIDENCE (1996); DAVID A. SCHUM, THE EVIDENTIAL FOUNDATIONS OF
PROBABILISTIC REASONING (1994).
5. For research on juries, see, for example, INSIDE THE JUROR: THE PSYCHOLOGY OF JUROR
DECISION MAKING (Reid Hastie ed., 1993).
6. See, e.g., Kurt M. Saunders, Law as Rhetoric, Rhetoric as Argument, 3 J. ASSN LEGAL
WRITING DIRECTORS 166, 166-67 (2006) (discussing law as consisting of pragmatic decisions
rather than formally logical ones, in which rhetoric, as opposed to formal logic, influences the
conclusions of legal reasoning); Mary Massaron Ross, A Basis for Legal Reasoning: Logic on
Appeal, 3 J. ASSN LEGAL WRITING DIRECTORS 179, 181 (2006) (arguing that despite the theory of
rhetoric as outcome-determinative of legal decision-making, formal logic can be a possible
rhetorical device and a critical element, especially at the appellate level).
7. For research on the Toulmin Model, see, for example, ARGUING ON THE TOULMIN
MODEL (David Hitchcock & Bart Verheij eds., 2006). For research based on argumentation
schemes, see DOUGLAS N. WALTON, ARGUMENT SCHEMES FOR PRESUMPTIVE REASONING 13-14,
46 (1996) (stating that the function of an argumentation scheme is to orchestrate a dialogue by use
of appropriate critical question[s], the asking of which shifts a burden or weight of presumption
to the other side in a dialogue). For examples of research on argumentation schemes within AI and
law, see Henry Prakken, Chris Reed, & Douglas Walton, Argumentation Schemes and
Generalizations in Reasoning About Evidence, 9 INTL CONF. ARTIFICIAL INTELLIGENCE & L.
PROC. 32 (2003); Henry Prakken & Giovanni Sartor, The Three Faces of Defeasibility in the Law,
17 RATIO JURIS 118, 120-23 (2004).
8. See, e.g., BART VERHEIJ, VIRTUAL ARGUMENTS: ON THE DESIGN OF ARGUMENT
ASSISTANTS FOR LAWYERS AND OTHER ARGUERS 97-122 (2005); Kevin D. Ashley & Edwina L.
Rissland, Law, Learning and Representation, 150 ARTIFICIAL INTELLIGENCE 17 (2003) [hereinafter
Ashley & Rissland, Law, Learning]; Edwina L. Rissland, Artificial Intelligence and Law: Stepping
Stones to a Model of Legal Reasoning, 99 YALE L.J. 1957 (1990) [hereinafter Rissland, Artificial
Intelligence]; Edwina L. Rissland, Kevin D. Ashley & R.P. Loui, AI and Law: A Fruitful Synergy,

2007]

DISCOVERING THE LOGIC OF LEGAL REASONING

1689

however, is a concerted effort within our profession to articulate the

general logic of our method of reasoning, and to do so in a way that is
useful in solving legal problems and which provides us a normative ideal
of sound reasoning.
If we had such a useful, normative method, it would surely be
evident in the classroom, and ultimately in the law office and courtroom.
Most legal educators, however, merely illustrate reasoning by exhibiting
a stream of examples (both historical and hypothetical), leaving it to the
student to abstract from those examples how to think like a lawyer.
We seldom develop general accounts that explain to students how
lawyers ought to reason and why. The closest we come to training
students in generalized methods of reasoning is probably in the context
of legal writing,9 and perhaps also in some instances of skills training.10
But these relatively recent developments have not penetrated very far
into the doctrinal core of the legal curriculum. We do not even instruct
students in deductive logic, whose relevance we surely acknowledge,11
perhaps because we recognize its limited usefulness in solving difficult
legal problems.12 Although traditional logic captures part of legal
reasoning, it does not address the most difficult parts, and so its study
may not be particularly critical for lawyers. If we had a truly useful
theory of legal reasoning, we would surely place it at the center of our
legal curriculum.
This Idea begins a discussion about why legal reasoning may
exhibit distinctive features that merit logical analysis. It suggests that the
demands of the rule of law combine with the pragmatic nature of legal
reasoning to evolve distinctive patterns of reasoning. The Idea briefly
150 ARTIFICIAL INTELLIGENCE 1 (2003).
9. For an excellent introduction to legal reasoning within the context of legal writing, see
RICHARD K. NEUMANN, JR., LEGAL REASONING AND LEGAL WRITING: STRUCTURE, STRATEGY,
AND STYLE (5th ed. 2005).
10. E.g., STEFAN H. KRIEGER & RICHARD K. NEUMANN, JR., ESSENTIAL LAWYERING SKILLS:
INTERVIEWING, COUNSELING, NEGOTIATION, AND PERSUASIVE FACT ANALYSIS 127-211 (2d ed.
2003) (discussing persuasive fact analysis).
11. TERENCE ANDERSON & WILLIAM TWINING, ANALYSIS OF EVIDENCE: HOW TO DO
THINGS WITH FACTS BASED ON WIGMORES SCIENCE OF JUDICIAL PROOF 6369 (1991) (showing
that for Wigmore the principle utility of the deductive form of inference was to force into
prominence the generalization upon which the inference rests); Mark L. Movsesian, Rediscovering
Williston, 62 WASH. & LEE L. REV. 207, 24143 (2005) (describing Willistons view that the use of
analytic logic in law has pedagogical benefits, promotes predictability and stability in law, and
makes the legal system more acceptable to the general public).
12. See, e.g., Leonard G. Boonin, The Logic of Legal Decisions, 75 ETHICS 179 (1965)
(discussing the deductive model of reasoning and the problem of explaining how to apply legal rules
in a given case, how to elaborate on rules, and how to create new rules); Jeffrie G. Murphy, Law
Logic, 77 ETHICS 193 (1967) (arguing that logical and legal validity are different and in some ways
incompatible).

1690

HOFSTRA LAW REVIEW

[Vol. 35:1687

discusses three types of legal reasoning. Rule-based reasoning and

evidence evaluation, as they are found in law, exhibit distinctive logical
features. So does second-order process reasoning, which can modify
both rule-based reasoning and evidence evaluation. Taken together,
these three types give legal reasoning a complex default character that
is distinctive to it. In addition, the structure of the legal community
promotes the evolution of reasoning patterns that are well-adapted to the
task of solving legal problems. Empirical research is needed to discover
the actual patterns that have evolved. This Idea cannot of course lay out
designs for such empirical research. It must be enough for now to
suggest why such research is needed, and why it promises to be a
successful means of discovering the logic of legal reasoning.
II.

THE PRAGMATIC NATURE OF LEGAL REASONING

The legal profession has not found traditional formal logic very
useful because that logic tracks the deductive reasoning of mathematics,
and not the practical reasoning we actually employ in law. We are a
pragmatic profession. We face an explosion of legal problems that
require our attention and we have developed our own techniques to help
us think through such problems. We therefore have little time for formal
theories that are not tailored to our mode of reasoning or do not make us
more effective or efficient at solving our professional problems. We will
not adopt any methodology that is not also useful.
The reasoning techniques we employ are pragmatic in at least three
senses. The first is that our reasoning is action-oriented. We use it to
make legal arguments about the legitimacy of governmental action or
inaction. When judges decide cases or administrative agencies adopt
new regulations, they must interpret constitutional, statutory, or
regulatory texts, and balance legal principles against substantive
policies. Before a court can lawfully sentence a defendant or enter a
binding judgment, or before an administrative agency can lawfully adopt
a rule or issue a permit, it must evaluate the legally available evidence.
Legal reasoning is therefore pragmatic in the sense that its ultimate
subject matter is governmental action, and is almost always about
justifying decisions leading to such action.
Second, legal reasoning is pragmatic in the sense that it balances
the epistemic objective of law against the applicable non-epistemic
objectives.13 The epistemic objective is to produce determinations of
fact that are as accurate as possible and which are warranted by the
13. Vern R. Walker, Epistemic and Non-Epistemic Aspects of the Factfinding Process in Law,
3 APA NEWSL., PHIL. & L., Fall 2003, at 132.

2007]

DISCOVERING THE LOGIC OF LEGAL REASONING

1691

legally available evidence. The epistemic side of law aims at truth, but a
truth constrained by reasonable inferences from the evidence. Weighed
against this epistemic objective are numerous non-epistemic objectives.
Some of them are common across governmental institutions and
proceedings (for example, procedural fairness or administrative
efficiency), while others are limited to particular institutions and
proceedings (for example, achieving an adequate supply of electric
power, or increasing economic efficiency within securities markets).
Each institution strikes its own peculiar balance of epistemic and nonepistemic objectives. Administrative agencies that have a mandate to
protect public health may design their factfinding processes to be as
thorough and accurate as possible, but consistent with ensuring the
protection of the public. Criminal courts should design their factfinding
processes to avoid erroneous verdicts, but must also protect the due
process rights of the accused and achieve reasonable levels of judicial
efficiency. For any particular institution and substantive mission, the
overall goal is to strike the appropriate balance between the epistemic
objective and the applicable non-epistemic objectives. Legal reasoning is
pragmatic because it must incorporate such balancing and must reason
about appropriate balancing.
Third, legal reasoning is pragmatic because legal decision-making
occurs in real time, uses limited resources, and is usually based on
incomplete information. Legal decision-making is decision-making
under uncertainty.14 The decision-maker has to evaluate, at each stage of
the process, whether the evidence is complete enough, whether the
residual uncertainty is acceptable, and whether action should be taken or
postponed. A prosecutor could always conduct more investigation and a
regulator could always obtain more scientific studies. When and how to
take which type of action involves an assessment of the current state of
uncertainty, of the costs and benefits of obtaining more evidence, and of
the risks and benefits of acting without additional evidence. Legal
reasoning is pragmatic because it must take into account such resource
considerations.
These three pragmatic dimensions of legal reasoning dictate certain
features of its logic. Legal reasoning evaluates decisions and actions,
balances epistemic and non-epistemic objectives, and occurs under the
constraints of limited resources and incomplete information. The logic
of legal reasoning must incorporate all of these dimensions. It would go

14. E.g., JUDGMENT UNDER UNCERTAINTY: HEURISTICS AND BIASES (David Kahneman, Paul
Slovic & Amos Tversky eds., 1982); M. GRANGER MORGAN & MAX HENRION, UNCERTAINTY: A
GUIDE TO DEALING WITH UNCERTAINTY IN QUANTITATIVE RISK AND POLICY ANALYSIS (1990).

1692

HOFSTRA LAW REVIEW

[Vol. 35:1687

well beyond deductive logic, and incorporate inductive and abductive

logic15 (what logicians today call nonmonotonic logic16). It would draw
upon modern decision theory, risk-benefit analysis, and risk analysis.
Moreover, to remain useful to the law, such methodologies and theories
would not be incorporated entirely or for their own sakes, but only in so
far as they improve legal decision-making and do not distort our
accepted patterns of reasoning.
It is no surprise then that the profession largely ignores formal
systems that are not necessary for solving legal problems. Although
logicians and researchers in artificial intelligence regard formal models
as benign and useful, legal theorists and practitioners generally regard
them with suspicion. Although we all accept the use of basic arithmetic
without question, legal decision-makers and stakeholders in the legal
process often require considerable explanation before they trust formal
systems such as probability theory and statistics. An elegant
mathematical proof may be perfectly transparent to a highly educated
mathematician, but it is useless as part of legal reasoning unless it is
transparent to legal decision-makers and to the parties. Introducing
formal complexity into legal reasoning can be inefficient, confusing, and
counterproductive, unless that added complexity is essential to solving
real legal problems.17 The same is true of formal logics. Formal theories
of legal reasoning are useful only if they accurately capture the
distinctive features of such reasoning, provide a useful normative
standard for evaluating such reasoning, and solve problems that cannot
be readily solved without those theories. Theories about the logic of
legal reasoning must be useful in order to be normative.
This pragmatic context therefore strongly influences what we are
looking for in the three types or areas of legal reasoning discussed next
in this Idea: rule-based reasoning, evidence evaluation, and second-order
reasoning about the decision-making process. In each area, the challenge
15. For an account of abductive reasoning, see John R. Josephson & Michael C. Tanner,
Conceptual Analysis of Abduction, in ABDUCTIVE INFERENCE: COMPUTATION, PHILOSOPHY,
TECHNOLOGY 5, 5 (John R. Josephson & Susan G. Josephson eds., 1994) (Abduction, or inference
to the best explanation, is a form of inference that goes from data describing something to a
hypothesis that best explains or accounts for the data.).
16. E.g., GERHARD BREWKA, JRGEN DIX & KURT KONOLIGE, NONMONOTONIC REASONING:
AN OVERVIEW (1997); HENRY E. KYBURG, JR. & CHOH MAN TENG, UNCERTAIN INFERENCE 117-51
(2001); ISAAC LEVI, FOR THE SAKE OF THE ARGUMENT 120-59 (1996); HENRY PRAKKEN, LOGICAL
TOOLS FOR MODELLING LEGAL ARGUMENT: A STUDY OF DEFEASIBLE REASONING IN LAW 67-100
(1997).
17. Formal systems, such as those employing Bayesian techniques, can create more problems
than they solve. See generally Vern R. Walker, Language, Meaning, and Warrant: An Essay on the
Use of Bayesian Probability Systems in Legal Factfinding, 39 JURIMETRICS 391 (1999) (book
review) (critiquing generally the use of formalizations in legal factfinding).

2007]

DISCOVERING THE LOGIC OF LEGAL REASONING

1693

is to make transparent those features of reasoning that are useful in

solving legal problems.
III.

RULE-BASED REASONING

A primary strategy for deciding similar cases similarly is to develop

and apply substantive legal rules, which prescribe particular outcomes
for particular types of cases. The substantive rules of law state the
conditions under which particular types of governmental action are
justified. Formal logic represents such rules as conditional
propositions.18 A proposition is the descriptive content of an assertion
or statement. It is capable of being either true or false, and is usually
expressed in ordinary language by a sentence or a clause. A
conditional proposition has the logical form if p, then q, where p and
q stand for two constituent propositions. In the terms of this conditional
schema, a legal rule states that if proposition p (the condition) is true,
then this fact warrants that proposition q (the conclusion) is also true. A
warranted conclusion can then warrant additional inferences, based on
additional rules, and can ultimately help justify action or inaction.
While traditional logic has focused on propositions as having one
of two values (true and false), the dynamic process of rule-based
legal reasoning is better understood as assigning to propositions one of
three values (true / undecided / false). When a legal proceeding
begins, all propositions that form the conditions of the applicable legal
rules are undecided. Participants in the legal process produce evidence
and arguments to persuade the decision-maker (whether judge, regulator,
or factfinder) to change the values of those propositions to either true
or false. Put another way, the legal rules identify the propositions that
are relevant within the type of proceeding, but the particular proceeding
begins with the decision-maker being neutral on whether the conditions
for applying those rules are satisfied or not.
A major feature of rule-based legal reasoning is the distinction
between prima facie case and affirmative defense. Some rules state the
conditions under which governmental action is justified (the prima facie
case for the proponent of the action), while other rules state exceptions
or affirmative defensesthat is, conditions under which the prima facie
line of reasoning is defeated. These latter rules, which logicians call
18. See Scott Brewer, Exemplary Reasoning: Semantics, Pragmatics, and the Rational Force
of Legal Argument by Analogy, 109 HARV. L. REV. 923, 972 (1996) (defining rule in a logically
spare manner, as a prescriptive proposition that has a logical structure the most abstract form of
which is reflected in the standard conditional proposition, either propositional (if P then Q) or
predicate (for all x, if x is an F then x is a G)).

1694

HOFSTRA LAW REVIEW

[Vol. 35:1687

defeaters,19 function as a kind of negation. If the defeater condition is

determined to be true, then the conclusion is false. Legal rules exhibit
defeater logic when they state an exception to a normal rule or an
affirmative defense to a prima facie case, and it is common in such
circumstances to place the burden of proof for the defeater proposition
on the party invoking the exception or raising the affirmative defense.
For example, the law of battery has rules governing when the defendant
has a privilege to act that constitutes an exception or defense to the
prima facie case (for example, when the defendant was acting
reasonably in making a lawful arrest or in defending herself from
intentionally inflicted bodily harm).
We can visualize systems of substantive rules of law (such as the
law of battery) as inverted implication treestrees that map the
conditions of rules, or the implications of proving issues of fact.20 Figure
1 shows a partial implication tree for the law of battery. The nodes of an
implication tree are propositions to be proved or disproved, and the top
node of a tree is the ultimate issue to be proved before some
governmental action is justified (for example, entering a court judgment
for the plaintiff). Each level of each branch extending downward from
the top node states the logical conditions for proving the immediately
higher proposition. Figure 1 illustrates both a conjunctive level (stating a
conjunction of conditions connected by AND) and a disjunctive level
(stating a disjunction of conditions connected by OR). A conjunction
is true if, but only if, all of the conjuncts are true, while a disjunction is
true if, but only if, at least one disjunct is true. A branch can also state a
defeating condition (connected by UNLESS), the truth of which
determines the conclusion to be false, even if the prima facie branch is
true. Because rules tend to have multiple conditions for making an
inference to a single conclusion, successively lower levels of a rule tree
tend to expand horizontally as they expand downward. The shape of the
inverted implication tree therefore tends to be triangular, with the single
ultimate issue as the apex at the top, dependent for its truth or falsehood
upon combinations of factual issues that terminate the branches along
the triangles base at the bottom.

19. For logical discussions of the defeater connective, see BREWKA ET AL., supra note 16, at
2-3, 16; JOHN L. POLLOCK, NOMIC PROBABILITY AND THE FOUNDATIONS OF INDUCTION 79 (1990).
For discussions in the field of AI and law, see Prakken et al., supra note 7, at 32, 37-38; Prakken &
Sartor, supra note 7, at 120-24; Henry Prakken & Giovanni Sartor, Reasoning with Precedents in a
Dialogue Game, 6 INTL CONF. ARTIFICIAL INTELLIGENCE & L. PROC. 1, 3 (1997); Vern R. Walker,
A Default-Logic Paradigm for Legal Fact-Finding, 47 JURIMETRICS 193, 199-204, 213-15 (2007).
20. For the use of tree structures in logic, artificial intelligence, and models of legal reasoning,
see Walker, supra note 19, at 201 & n.35.

2007]

DISCOVERING THE LOGIC OF LEGAL REASONING

Figure 1.

1695

Partial Implication Tree for the Law of Battery

Traditional logicians are used to the deductive logic of

mathematics, in which proof begins with axioms (at the top) and
deduces conclusions (downward). They may therefore overlook this
fundamentally different orientation of legal reasoning. The rule-based
deductions of legal reasoning do not rest upon self-evident axioms, but
rather upon the truth or falsehood of the relevant issues of fact, which
are defined by rules adopted by legal authorities. The soundness of the
reasoning depends upon the appropriateness of the rules governing the
reasoning, and (at bottom) upon the plausibility of the relevant evidence.
Moreover, conclusions that rest upon uncertain and incomplete evidence
are at best plausible and conditionally true. The rule-based reasoning of
law can be only as sound as the factfinders evaluation of the evidence
(which is the topic of the next section of this Idea).
Another feature of rule-based legal reasoning that challenges
traditional deductive logic is the possibility of changing the rules
themselves as a result of the reasoning. For example, within common
law systems, courts have inherent authority to elaborate new legal rules
that apply to the very case being decided, as well as to future cases. Even
when the authority is legislative and the legal rules are derived from
statutes or regulations, a court or administrative agency has considerable
discretion to elaborate new rules of application in pending cases.
Whenever a court or agency explicates a new definition for a legal term,

1696

HOFSTRA LAW REVIEW

[Vol. 35:1687

or interprets a legal phrase, or carves out an exception to an existing

rule, it creates a new rule. Such new rules may create new conditions
that extend the branches of the rules tree, or create exceptions that add
defeaters to the tree. Courts sometimes also overrule prior cases, hold
statutes unconstitutional, or vacate administrative regulations, thus
removing branches from the tree. Under the rule of law, however, the
action of changing a substantive rule is itself governed by legal rules and
must be justified in each particular case. This second-order aspect of
legal reasoning (reasoning about the rules themselves) will be discussed
in a later section of this Idea.
IV.

EVIDENCE EVALUATION

The legal rules therefore identify those issues of fact that are
relevant to proving the ultimate issue of fact. As the branches of the
implication tree extend downward, the terminal conditions at the end of
each sub-branch (the last propositions in each chain) constitute the issues
of fact that are relevant. In any particular case, various participants (such
as private parties, prosecutors, or administrative staffs) produce evidence
for the legal record, and use that evidence to try to prove or disprove
those issues of fact. The factfinders role is evidence evaluation:
deciding which evidence is relevant to which issues of fact, evaluating
the probative value of the relevant evidence, and making findings of fact
based on that evidence. The logic of evidence evaluation, therefore,
studies the methods and principles for the inferential aspects of the
factfinders task. It explains the reasoning that a reasonable factfinder
would use to determine the probative value of the evidence.
As in the case of rules, a basic building block of evidence
evaluation is the proposition. The propositions that constitute the
evidence can be called evidentiary assertions, or simply assertions,
to distinguish them from the propositions that constitute rules. Examples
of evidentiary assertions are statements made by testifying witnesses or
statements contained in documents that are admitted into evidence.
Although witnesses and documents provide many of the evidentiary
assertions in a particular case, the factfinder formulates additional
assertions that play a role in the reasoningfor example, a description
of an evidentiary exhibit or of a witnesss demeanor. While legal
authorities create the rules for categories of cases, the witnesses, the
documents, and the factfinder create the evidentiary assertions in each
particular case.
Although evidentiary assertions are propositions, and therefore
capable of being either true or false, factfinders are permitted to assign

2007]

DISCOVERING THE LOGIC OF LEGAL REASONING

1697

them degrees of plausibility (plausibility-values) instead of truthvalues.21 In evaluating any particular evidentiary assertion, the legal
rules also allow the factfinder to select the scale of plausibility to use.
For example, the factfinder might use an ordinal, five-valued plausibility
scale for some evidentiary assertions (with values such as highly
plausible / somewhat plausible / undecided / somewhat
implausible / highly implausible) and an ordinal, seven-valued scale
for other assertions (with values such as highly plausible / very
plausible / slightly plausible / undecided / slightly implausible /
very implausible / highly implausible). On rare occasions,
circumstances might even warrant using mathematical probabilities,
with real numbers between zero and one as possible values.
A reasonable factfinder would select a scale of plausibility that is
appropriately precise, given both the reliability of the available evidence
and the accuracy and reliability needed to make the relevant findings of
fact. For example, a factfinder might have little confidence in assessing
the credibility of a particular fact witness, and so evaluate that witnesss
assertions on an ordinal scale having very few degrees of plausibility.
Occasionally, however, an expert might persuade the factfinder that a
causal system is so well understood (for example, DNA profiling) that
the factfinder can evaluate assertions about it using mathematical
probabilities. On the other hand, such precision may not be necessary to
perform the factfinding task. In a legal proceeding whose findings are
made by the preponderance standard of proof, the evaluative precision
needed may be rather low, and an ordinal scale with a small number of
values may be entirely adequate. The beyond-a-reasonable-doubt
standard, however, probably requires greater evaluative precision in
order to support a verdict.
Evidence evaluation poses some challenges for formulating a
useful, normative logic. One challenge is explaining how we reason
about relevance itselfhow we decide to link particular evidentiary
assertions to particular issues of fact. Another challenge is studying the
logical properties of different plausibility scales, although we can learn
lessons from the theories of scientific classification and measurement.22
21. See id. at 209-12.
22. For general texts on measurement theory, see, for example, EDWARD G. CARMINES &
RICHARD A. ZELLER, RELIABILITY AND VALIDITY ASSESSMENT (1979); EDWIN E. GHISELLI, JOHN
P. CAMPBELL & SHELDON ZEDECK, MEASUREMENT THEORY FOR THE BEHAVIORAL SCIENCES
(1981); MEASUREMENT ERRORS IN SURVEYS (Paul P. Biemer et al. eds., 1991). For general statistics
texts with good treatments of measurement error, see, for example, DAVID FREEDMAN, ROBERT
PISANI, ROGER PURVES & ANI ADHIKARI, STATISTICS (2d ed. 1991); HERMAN J. LOETHER &
DONALD G. MCTAVISH, DESCRIPTIVE AND INFERENTIAL STATISTICS: AN INTRODUCTION (4th ed.
1993). For discussion within a legal context, see David H. Kaye & David A. Freedman, Reference

1698

HOFSTRA LAW REVIEW

[Vol. 35:1687

A further challenge is combining the plausibility-values of numerous

evidentiary assertions into a single plausibility-value for a particular
conclusion. This is a complicated problem in law because factfinders
must be able to integrate both non-expert and expert evidence into a
single pattern of reasoning. The warrant for the non-expert evidence
might be commonsense reasoning about an eyewitnesss perceptual
abilities and credibility, while the warrant for the scientific evidence
might be controlled laboratory experiments and epidemiological studies.
In the end, the factfinder must reason from all of the relevant evidence to
a particular finding of fact.
Ideally, we want to identify the patterns of default reasoning that
are actually used in law. If such patterns could be formalized into
plausibility schemas,23 they might furnish useful, normative models
for factfinders. In logic, a schema is a formal linguistic pattern
containing variables, so that appropriate substitutions for the variables
create instances of the pattern.24 A plausibility schema is a pattern of
default reasoning that, when instantiated, warrants the conclusion to be
plausible. For example, the schemas of deductive logic (such as modus
ponens25), which necessarily preserve truth from premises to
conclusion, also preserve plausibility from premises to conclusion. More
useful in typical legal cases, however, would be schemas based on
inductive or abductive logic, or derived from scientific methodologies or
heuristics.26 The typical plausibility schema has an inverted tree
structure similar to that of an implication tree, except that the logical
operators connecting the lower-level assertions to the upper-level
Guide on Statistics, in REFERENCE MANUAL ON SCIENTIFIC EVIDENCE 83, 102-04 (Federal Judicial
Center, 2d ed. 2000); Vern R. Walker, The Siren Songs of Science: Toward a Taxonomy of
Scientific Uncertainty for Decisionmakers, 23 CONN. L. REV. 567, 580-88 (1991).
23. For a discussion of plausibility schemas, see Walker, supra note 19, at 212-18.
24. For the use of schemas in logic to specify sets of permissible axioms or inferences, see,
for example, JOHN M. ANDERSON & HENRY W. JOHNSTONE, JR., NATURAL DEDUCTION: THE
LOGICAL BASIS OF AXIOM SYSTEMS 20-21 (1962); GERALD J. MASSEY, UNDERSTANDING
SYMBOLIC LOGIC 139-40, 147-49 (1970). For the use of schemas in semantics to specify conditions
for assigning a truth-value to a sentence, see, for example, JOHN I. SAEED, SEMANTICS 89, 305-09
(2d ed. 2003).
25. This schema has the form If p then q; p; therefore, q. COPI & COHEN, supra note 1, at
324, 373-74.
26. An example of a plausibility schema important in legal factfinding is what logicians call a
statistical syllogism. A statistical syllogism draws a conclusion about a specific individual based
on a statistical generalization (such as X percent of As are B). See JOHN L. POLLOCK & JOSEPH
CRUZ, CONTEMPORARY THEORIES OF KNOWLEDGE 229-30 (2d ed. 1999); WESLEY C. SALMON,
LOGIC 87-91 (2d ed. 1973); Josephson & Tanner, supra note 15, at 23. Toulmin has referred to such
inferences as quasi-syllogisms. See STEPHEN TOULMIN, THE USES OF ARGUMENT 108-11, 131-34,
139-40 (1958). For early recognition of the difficulty such inferences pose for legal theory, see
generally George F. James, Relevancy, Probability, and the Law, 29 CAL. L. REV. 689 (1941).

2007]

DISCOVERING THE LOGIC OF LEGAL REASONING

1699

conclusion function on the plausibility-values of the evidence, instead of

on the truth-conditions of rules. Such schemas formulate patterns of
plausible reasoning that warrant drawing provisional conclusions, which
then remain subject to re-evaluation. Identifying those plausibility
schemas that are actually used in legal factfinding will require empirical
research into the reasoning patterns of judges, administrative officers,
and expert agencies.27
When a legal proceeding begins, the applicable legal rules identify
all of the issues of fact that may be relevant. The factfinder then links the
legally available evidentiary assertions to those issues of fact, using as
heuristics those patterns of default reasoning familiar to the factfinder.
(A logic of evidence evaluation would try to capture the acceptable
patterns of reasoning as plausibility schemas.) The choice of pattern
depends upon the nature of the issue of fact to be proved and the nature
of the available evidence. When patterns of evidence are linked or
attached to the terminal propositions of the inverted rule tree, they
extend the branches of that tree further downward. As the schematic in
Figure 2 suggests, the complete logical model for the legal reasoning in
a particular case (the inference tree for the case) has the shape of an
inverted triangle, with the implication tree generating the upper branches
of the triangle and the attached patterns of evidence evaluation extending
those branches downward to the evidentiary assertions.
Figure 2.

General Structure of Reasoning in a Legal Case

27. For two examples of analyzing the reasoning of special masters in factfinding under the
National Childhood Vaccine Injury Act of 1986, see Walker, supra note 19, at 226-32; Vern R.
Walker, Visualizing the Dynamics around the Rule-Evidence Interface in Legal Reasoning, LAW,
PROBABILITY & RISK (forthcoming 2007), available at http://lpr.oxfordjournals.org/cgi/reprint/
mgm015?ijkey=GIBiZmvwOPjF0l6&keytype=ref.

1700

HOFSTRA LAW REVIEW

[Vol. 35:1687

After the factfinder organizes the evidence by relevance, evaluation

can proceed upward from the bottom of the extended tree. After the
factfinder assigns plausibility-values to the evidentiary assertions, the
instantiated plausibility schemas warrant the plausibility-values of
higher-level assertions. At the point in each branch where the
evidentiary assertions end and the condition of a rule occurs, the
applicable standard of proof directs the factfinder about how to make a
finding of fact based on the plausibility of the evidence. A
preponderance-of-evidence standard is the rule that a factfinder must
find the issue of fact to be true if the totality of relevant evidence is to
any degree plausible, and must find the issue of fact to be false if that
evidence is to any degree implausible. In addition, the legal rules on
burden of persuasion determine which party must lose (which finding to
make) if the evidence is undecided or in equipoise. Discovering the
logic of legal reasoning means making this process of evidence
evaluation transparent.
Devising a useful, normative logic of factfinding is even more
difficult because of the pragmatic and dynamic nature of the factfinding
process. Factfinders must make findings in real time, using limited
resources, and on the basis of the incomplete evidence. Moreover, the
factfinding process is often highly structured around the factfinder, with
participants other than the factfinder deciding what the legal rules are
and producing evidence on the record. As a result, all conclusions are
usually only tentative and subject to revision. New evidence, or a reanalysis of old evidence, can defeat an earlier conclusion or undermine
its evidentiary support. A change in rules may modify what lines of

2007]

DISCOVERING THE LOGIC OF LEGAL REASONING

1701

reasoning are acceptable. Yet at each stage of the legal proceeding, it

must be reasonable to rely upon each provisional conclusion. These four
characteristics of evidence evaluationthat it is practical, dynamic,
defeasible, and presumptively soundmake it an exemplary instance of
what logicians call default reasoning. Default reasoning uses the
available evidence, together with default inference rules, to warrant
presumptive conclusions, which are then subject to future revision.28 The
highly structured framework in which law accomplishes evidence
evaluation poses significant challenges to traditional logic.
V.

SECOND-ORDER PROCESS REASONING

Legal decision-making is itself a process governed by the rule of

law, and a third area of legal reasoning warrants conclusions about the
structure of that process. Process rules allow the decision-making
process to be dynamic, participatory, and interactive, while ensuring that
the process serves the rule of law and the appropriate balance of
epistemic and non-epistemic objectives. Different participants can play
different roles, with divisions of labor and responsibility, ideally within
a single, fair, and efficient process. Some participants have authority to
constrain the decision-making power or discretion of other participants.
For example, parties, trial judges, juries, and appellate judges have
distinct roles in judicial trials, and public commenters, regulators, and
reviewing courts have other roles in administrative rulemakings.
Proceedings consist of many points where different participants must
make decisions, and many of those decisions (such as rulings on
motions) are themselves actions governed by legal rules and must be
warranted by the available evidence.
From the standpoint of logic, such process decisions involve the
same kind of rule-based reasoning and evidence evaluation discussed
above. The ultimate issue to be decided may be whether the court has
jurisdiction to adjudicate a particular case, or whether a particular
document is admissible as evidence. For such decisions, legal rules
define terms and structure the acceptable lines of reasoning, and the
issues of fact often require evidence evaluation. For example, the
citizenship of a party might be a factor in deciding jurisdiction, or the
method of obtaining a document might affect its admissibility as
evidence. The logic of rule-based reasoning and evidence evaluation,
therefore, also applies to the reasoning about process decisions.
It is traditional to divide process decisions, as well as the rules

28. See Walker, supra note 19, at 194-95.

1702

HOFSTRA LAW REVIEW

[Vol. 35:1687

governing those decisions, into two types: procedural and evidentiary.29

Procedural rules address issues as general as jurisdiction, or as specific
as the appropriateness of particular filings. Procedural decisions based
on those rules orchestrate the dynamics and timing of the decisionmaking process. Evidentiary rules address issues about the evaluation of
evidence, such as the admissibility of evidence, the legal sufficiency of
evidence, and burdens and standards of proof. Evidentiary decisions
based on those rules manage the various tasks involved in evidence
evaluation, and allocate them to various participants in the process. For
example, the judge decides whether proffered evidence is admissible,
while the factfinder decides the probative value of that evidence once it
has been admitted. Usually, the same procedural and evidentiary rules
apply across many types of casesfor example, rules of criminal
procedure and evidence apply to criminal cases generally, while
administrative procedure statutes apply to administrative proceedings
generally.
Reasoned decision-making about the very process of decisionmaking has several distinctive features that create challenges for a
useful, normative logic. One is that process decisions often involve
second-order reasoning. That is, the propositions of process reasoning
are about other propositionsabout the propositions stating substantive
legal rules or about the evidentiary assertions in a particular case. For
example, a motion to dismiss a complaint for failure to state a cause of
action questions whether there is any legal rule that allows a judgment
favorable to the plaintiff. The motion claims that there is, in that
jurisdiction, no legal rule that would let the plaintiff succeed, even if the
plaintiff actually proves all of the plaintiffs allegations. A motion to
exclude a particular experts opinion from the evidence in a case
requires the court to reason about the experts proffered evidentiary
assertion. The fact that the rule of law applies also to decisions about the
legal process means that we need to have sound reasoning about
acceptable rule-based reasoning and about adequate evidence
evaluation. We need sound reasoning about legal reasoning.
An important type of process reasoning is policy-based reasoning
about whether to adopt, modify, or rescind legal rules. Motions may
raise questions about the interpretation of a statutes language, or about
creating an exception to a previously adopted substantive rule of law, or
about overruling a precedent. In the terminology introduced earlier in
this Idea, such reasoning addresses the shape of the implication tree
29. For extended examples of process decisions in the context of default reasoning, see id. at
232-41.

2007]

DISCOVERING THE LOGIC OF LEGAL REASONING

1703

itself. The rule of law requires that those decisions be justified in turn,
and we expect courts and administrative agencies to provide reasons for
adopting or modifying rules of law. Such justification involves
balancing the epistemic objective against the relevant non-epistemic
governmental objectives, with due regard for overriding legal principles
such as non-discrimination and due process. A logic of policy-based
reasoning would incorporate the content of policies and principles, and
would formulate methods for weighing many divergent lines of
reasoning in warranting decisions about particular rules.
Another area is analogical reasoning about case precedents.30
Changing legal rules incrementally is a practice that addresses concerns
for deciding similar cases similarly over time, for maintaining
predictability of outcome, and for providing due notice to potentially
affected parties. Decisions about adopting or modifying legal rules,
therefore, generally consider cases that were decided earlier. The logic
of legal reasoning should capture the kinds of reasons that courts
routinely give for considering two cases to be similar, and for
distinguishing one case from another. Making such reasoning
transparent will be a very difficult task. At a minimum, it will involve
identifying the attributes that are relevant for comparing legal cases,
devising a valid and reliable method of classifying actual cases on those
attributes, and determining how judges and regulators should decide
whether two cases are sufficiently similar or dissimilar.31 Any adequate
theory would take into account the pragmatic contexts of different cases,
including the balance of epistemic and non-epistemic policies that might
explain dissimilarities among cases.
There are undoubtedly other features of second-order process
reasoning that pose challenges to discovering its peculiar logic.
Accomplishing the task requires solid empirical research into how the
legal profession actually performs such reasoning, in both normal and
borderline cases. The final section of this Idea discusses why empirical

30. For general legal discussions of the problem of analogical reasoning about precedents,
see, for example, EDWARD H. LEVI, AN INTRODUCTION TO LEGAL REASONING (1949); Brewer,
supra note 18; Cass R. Sunstein, Commentary, On Analogical Reasoning, 106 HARV. L. REV. 741
(1993).
31. For research in Artificial Intelligence and law dealing with case-based reasoning, see, for
example, L. KARL BRANTING, REASONING WITH RULES AND PRECEDENTS: A COMPUTATIONAL
MODEL OF LEGAL ANALYSIS 6 (2000) (discussing the research goal of using rule-based reasoning
and case-based reasoning as complementary processes for classification and explanation in legal
analysis); Ashley & Rissland, Law, Learning, supra note 8, at 3354 (surveying the HYPO family
of case-based reasoning (CBR) models, including HYPO, CABARET, and CATO); Rissland,
Artificial Intelligence, supra note 8, at 196878 (surveying early developments in AI and law that
used case-based reasoning).

1704

HOFSTRA LAW REVIEW

[Vol. 35:1687

research into our actual reasoning is a necessary step in discovering a

useful logic for legal reasoning.
VI.

LANGUAGE, EVOLUTION, AND LOGIC

Legal reasoning is the method by which lawyers invent arguments,

judges and regulators make considered legal decisions, and students and
professionals learn the law. The professional community has developed
the language, concepts and patterns of reasoning that help us perform
these tasks well. The philosopher Ludwig Wittgenstein called such
linguistic practices forms of life or language games.32 Legal rules
show a propositions use by stating the conditions under which we
accept it as true, and implicitly show the meaning of legal concepts.
Thinking like a lawyer means knowing how to use legal concepts to
formulate those lines of reasoning that are acceptable to lawyers.
Practicing law well means using such reasoning effectively to influence
legal decision-makers. The logic of legal reasoning studies the patterns
of legal language that legal professionals generally regard as appropriate.
The study of legal logic is the study of the appropriate use of legal
language.
Legal language and reasoning, moreover, are evolving, as we adapt
them to solve new legal problems. The existence of a hierarchy of legal
decision-makers ensures that all legal professionals (those who seek to
influence the decisions of those decision-makers) use legal language in
ways that judges, regulators, and other attorneys all understand. Lawyers
want to speak coherently and effectively to judges, trial judges want to
give jury instructions that correctly state the law, and trial judges and
regulators want to give reasons that satisfy appellate and reviewing
judges. Legal language is not a haphazard affair, but is highly
disciplined and difficult to learn (as first-year law students can attest).
Moreover, many decision-makers document their legal reasoning in
written opinions and rulings. This documentation helps ensure continuity
of language and reasoning patterns over time. This broad communal
effort at solving recurring types of legal problems causes legal language
and reasoning to evolve into those patterns that the community finds
32. Understanding the meaning of a linguistic expression involves knowing how to use it
appropriately within the rich contexts of human activity that Wittgenstein called forms of life.
See, e.g., LUDWIG WITTGENSTEIN, PHILOSOPHICAL INVESTIGATIONS 4, 7, 10, 23, 27-29, 34-35, 6871, 116-17, 127-28, 192 (G.E.M. Anscombe trans., 3d ed. 2001). In particular, knowing how to use
a descriptive expression correctly requires knowing the conditions under which we consider it to be
true or false, at least in paradigmatic circumstances. Id. at 21-22, 24, 29-30, 48, 75, 99, 142;
LUDWIG WITTGENSTEIN, ON CERTAINTY 2, 8-9, 10-12, 28, 30 (G.E.M. Anscombe & G.H. von
Wright eds., Denis Paul & G.E.M. Anscombe trans., 1969).

2007]

DISCOVERING THE LOGIC OF LEGAL REASONING

1705

effective at solving such problems. The forces of evolution behind those

patterns make it likely that whatever we find will be both useful and
normative.
Discovering the evolving language and reasoning of law therefore
requires empirical research into the actual balances that have been struck
between epistemic and non-epistemic objectives in different pragmatic
contexts. The rule of law promotes evolution through an insistence that
similar cases be decided similarly, that decisions be based on transparent
evidence and reasoning, and that decision-making proceedings be
governed by process rules. Each particular area of law evolves some
concepts and modes of reasoning that are precisely tailored to achieving
its own balance of objectives. Only empirical research into the actual
reasoning behind actual decisions can discover the tailored patterns that
have evolved. But in addition, legal reasoning also employs certain
distinctive methods wherever it is foundthose methods that have
proved successful in achieving the epistemic objective within many legal
contexts. Such general methods include rule-based reasoning, evidence
evaluation, and second-order process reasoning. Even these general
methods require empirical research, if we hope to discover a useful logic
of legal reasoning.
For example, we need empirical research on how legal rules evolve
over time. One hypothesis is about how factfinding can result in new
legal rules. When a factfinder evaluates evidence and makes findings of
fact, and the proceeding is one in which the factfinders reasoning is
explicitly documented and carefully reviewed, then soft rules can
emerge.33 Soft rules in this context are patterns of reasoning that have
been publicly scrutinized and upheld on review as being reasonable. Soft
rules are safe havens that show later factfinders how evidence can be
reasonably evaluated, with likely immunity from second-guessing by
reviewing judges. After a consensus forms around soft rules as accepted
patterns of reasoning, however, courts may eventually harden such
patterns into conventional rules of law. This hypothesis is an example of
how empirical research may uncover mechanisms by which new rules
can evolve from past legal proceedings.
Empirical research on the evolution of evidence evaluation
promises even more progress, in part because so little empirical research
has been done. We should begin by discovering more about the patterns
of reasoning that factfinders in different areas actually find plausible,
and how those factfinders integrate non-expert and expert evidence into
a single conclusion or finding. We should also clarify how rule-based
33. For a discussion of this hypothesis, see Walker, supra note 27.

1706

HOFSTRA LAW REVIEW

[Vol. 35:1687

reasoning and evidence evaluation actually interact in particular legal

areasfor example, the interaction between decisions about
admissibility of evidence, or about the legal sufficiency of evidence, and
the evaluation of probative value. We might then discover how secondorder reasoning about evidence evaluation can evolve new patterns of
plausible reasoning. For example, rulings about the admissibility or
sufficiency of evidence may evolve new plausibility schemas for
evidence evaluation.
Finally, the evolutionary effect of second-order reasoning should
become an important area of empirical research. It is probably in this
area that the logic of legal reasoning is most distinctive. The legal
community has evolved rule-based reasoning and evidence evaluation as
a paradigm for substantive decision-making, but employs that same
paradigm in making decisions about the decision-making process itself.
This creates the possibility of a reasoned evolution of legal reasoning.
Empirically studying the mechanisms by which legal language evolves
holds the promise of also discovering its normative and useful logic.
VII.

CONCLUSION

The goal of this Idea is to be suggestive but open-ended about

discovering the logic of legal reasoning. These brief thoughts about the
rule of law and the pragmatic nature of legal reasoning, as well as about
rule-based reasoning, evidence evaluation, and second-order process
reasoning, can only suggest that there is something distinctive about
legal reasoning. The added thoughts about how legal language and
reasoning patterns evolve are intended to spur a desire for empirical
research into actual reasoning patterns, even if this Idea cannot propose
particular designs for carrying out such research. But we in the legal
profession have little incentive to engage in such research if we do not
sense a need for it. And if we do not sense that the reasoning we apply to
legal problems has any distinctive and coherent structure, we are
unlikely to try to study that structure. What we need is a professional
awakeningto the possibility of discovering a useful logic of legal
reasoning.