The Yale Law Journal Company, Inc.

Scientific Expert Testimony and Intellectual Due Process

Author(s): Scott Brewer
Source: The Yale Law Journal, Vol. 107, No. 6 (Apr., 1998), pp. 1535-1681
Published by: The Yale Law Journal Company, Inc.
Stable URL: http://www.jstor.org/stable/797336 .
Accessed: 26/01/2015 18:46

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].

The Yale Law Journal Company, Inc. is collaborating with JSTOR to digitize, preserve and extend access to
The Yale Law Journal.

http://www.jstor.org

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
Articles

Scientific Expert Testimony and

Intellectual Due Process

Scott Brewert

CONTENTS

I. THE DOCTRINALAND INSTITUTIONALFRAMEWORKOF

EPISTEMICDEFERENCE ............. ................. 1540
A. Evidence, Procedure, and the "Law's Epistemology" .... .. 1540
B. Dauberts Philosophy of Science ..................... 1542
C. Brown's Empirical Psychology ...................... 1553

II. AxIOLOGY AND "POINT OF VIEW" IN THE ANALYSIS OF

EPISTEMICDEFERENCE ............. ................. 1566
A. Central Relation: Epistemic Deference by "Practical" (Legal)
Reasoners to "Theoretical" (Scientific) Experts .... ...... 1566
B. Reasoning from a "Point of View" Pp. .................. 1568
C. The Practical and Theoretical as Points of View .... ..... 1576

III. "THEORETICAL"AUTHORITYAND EPISTEMICDEFERENCE ...... 1582

A. Believing a Person Versus Believing a Proposition .... .... 1582
B. Actual Versus Putative Authoritativeness .... ........... 1585
C. The Concept of Epistemic Deference ..... ............. 1586

t Professor,HarvardLaw School. I would like to repay debts of thanksfor various kinds of help,
rangingfrom brainstormingconversationsto detailedcriticismof drafts,to JonathanAmsterdam,Anthony
Appiah,BrianBix, Julius Christensen,RichardCraswell,CatherineElgin, CharlesFried,ElizabethGarrett,
Kent Greenawalt,Steven Gross, Deborah Hellman, Larry Lessig, Peter Murray,David Nagle, Martha
Nussbaum, and Cass Sunstein. I presentedearlier versions of this Article in legal theory workshops at
Boston University, Columbia University, Harvard University, McGill University, the University of
Californiaat Berkeley,the Universityof Chicago, and the Universityof Pennsylvania,and I am indebted
to workshopparticipantsfor valuablecomments on those versions.

1535

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1536 The Yale Law Journal [Vol. 107: 1535

IV. EXPERTISE:CONCEPTSAND BASIC PROBLEMSFOR THE NONEXPERT1590

A. Knowledge, Warranted Beliefs, and Degrees of
Epistemic Competence ........................... 1590
B. Practical Epistemic Deference and Theoretical Judgment ... 1593
C. The Nonexpert's Selection and Competition Problems ...... 1593

V. KNOWLEDGEVERSUS JUSTIFIEDBELIEF: WHAT Do NONEXPERTS

WANT FROMEXPERTS? .............................. 1596
A. What the Law Desires: Acquiring Justified Beliefs
from Experts ................. ............. . 1596
B. Testimony as a Source of Justified Belief ............... 1601
1. Testimony ............... .................. 1602
2. Expert Testimony ............................ 1608
C. Point: The Four Possible Routes of WarrantedEpistemic
Deference by Nonexperts to Experts ........ .......... 1616
1. First Route: Substantive Second-Guessing in Practical
Epistemic Deference ......... .. ............... 1616
2. Second Route: Using General Canons of Rational
Evidentiary Support .......................... 1618
3. Third Route: Evaluating Demeanor in Practical
Epistemic Deference .......................... 1622
4. Fourth Route: Evaluating Credentials ...... ........ 1624
a. The Regress Problem ..... ................. 1625
b. The Question-Begging Problem .... ........... 1627
c. The Underdetermination Problem .... .......... 1630
D. Counterpoint: The Anti-Skeptical Response and the
Dialectical Impasse .................. ........... 1630

VI. BREAKING THROUGHTHE DIALECTICALIMPASSE:A MODEL OF

RATIONALEPISTEMICDEFERENCETO EXPERTS ............. 1634
A. The Concept of "Practical Priority" .......... .......... 1634
B. "Practical Priority" in the Minor Premise of a
Practical Syllogism .............. ................ 1636
C. A Simplified Model of Practical Epistemic Deference ...... 1646
1. Summary Presentation of the Model .... ........... 1647
2. Comment on Step (1) ...... ................... 1647
3. Comment on Steps (2) to (10) ..... .............. 1656
D. Oversimplification in the Foregoing Model of Practical
Epistemic Deference ............................. 1657
E. Breaking the Dialectical Impasse and Completing the Model:
Abduction in Legal Reasoning ......... ............. 1658

VII. INTELLECTUALDUE PROCESS ......................... 1672

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific ExpertTestimony 1537

A. Epistemic Nonarbitrariness as a Practical Constraint on

Legitimate Epistemic Deference ...... ............... 1672
B. Consequences for Doctrinal and Institutional Design:
"Two-Hat" Solutions and Intellectual Due Process .... ... 1677

VIII. CONCLUSION............ ........................ 1679

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1538 The Yale Law Journal [Vol. 107: 1535

Suppose that two groups of expert mathematicians disagree about a

complex mathematical question-say, whether Princeton mathematician
Andrew Wiles really did solve "Fermat's Last Theorem," which no
mathematician had been able to prove since Louis Fermat first propounded it
about 360 years ago.' These experts have had an opportunity to hear one
another's reasons for their competing conclusions about Wiles's proof, and
neither group is convinced by the other. How might we decide which of the
two groups is making the correct mathematical judgment? Here's a suggestion:
Convene a group of twelve or so nonmathematicians, give them an opportunity
to hear from representatives of each of the competing groups of
mathematicians, and have the nonmathematicians decide whether Wiles's proof
really succeeded. If the truth of the matter was among one's chief concerns,
would this decision procedure seem sound? There is serious reason to doubt
it. The most obvious problem with such a procedure is that it seems to turn the
decision about this disputed, highly complex question in the science of
mathematics over to those who are least competent to answer it.
Many legal systems, including the state and federal systems of the United
States, use decision procedures that are disturbingly close to the one just
imagined, procedures in which nonexpert judges and juries are called upon and
authorized to evaluate expert scientific testimony. This Article's goal is to offer
a sustained critical analysis of the legal rules and doctrines that create and
administer this procedure. Expert scientific information is relevant to, even
decisively important in, a rapidly growing percentage of decisions throughout
civil and criminal law. Most judges and juries, however, are not sufficiently
familiar with relevant scientific fields to be able independently and reliably to
bring scientific information to bear on their decisions. Instead, they must solicit
and defer to the judgments of expert scientific witnesses.
Moreover, almost inevitably in litigated cases in which expert scientific
evidence is offered, nonexpert judges and juries are presented, not with one
authoritative "voice" of scientific truth, but instead with competing scientific
expert witnesses who testify to contrary or even contradictory scientific
propositions. Lacking the information necessary to make cogent independent
judgments about which of the competing scientific experts to believe,
nonexpert legal decisionmakers choose among the experts by relying on such
indicia of expertise as credentials, reputation, and demeanor. Thus, even the
act of soliciting and deferring to expert scientific judgment requires nonexperts
to use a reasoning process-the process of selecting the experts, deciding
which expert to believe when the experts compete, and, finally, deciding how

1. Accordingto this theorem,whenevern is greaterthan 2, the equationa' + b' = cncannotbe solved

in whole numbers.Wiles's proof is nearly 200 pages long. See SIMON SINGH, FERMAT'SENIGMA256
(1997) (discussing the first efforts of the mathematicalcommunityto check Wiles's proof).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1539

to use the believed expert's information in resolving the central dispute being
litigated.
Drawing on work in jurisprudence, epistemology, philosophy of science,
and theories of practical reasoning, as well as on doctrines and leading cases
on scientific expert evidence, this Article carefully models the reasoning
process by which nonexpert legal reasoners defer to scientific experts in the
course of applying a law to individual litigants. Drawing on this model, I argue
for four central conclusions. Taken together, these conclusions have far-
reaching consequences for virtually all legal systems in which nonexpert legal
decisionmakers confront expert scientific testimony.
First, the Article argues that in order to avoid making an epistemically
arbitrary choice about which of the competing scientific experts ought to be
believed, a person must understand (in a special sense discussed in the text)
the cognitive aims and methods of science. But nonexpert judges and juries
lack just that kind of understanding, which is why they rely instead on other
indicia of expertise, such as credentials, reputation, and demeanor. Second,
nonexpert judges' and juries' lack of understanding of the cognitive aims and
methods of science and their reliance on such indicia of expertise as
credentials, reputation, and demeanor to choose between competing scientific
experts thus yield only epistemically arbitrary judgments. Third, the
conclusions that nonexpert judges and juries ultimately reach by relying
significantly on expert scientific testimony are often also epistemically arbitrary
and are therefore not justified from a legal point of view. Fourth, I
identify-and begin the process of explicating-an emerging norm that belongs
in the family of rule-of-law values: the norm I call intellectual due process.
Showing that this norm is immanent in values to which many legal
systems-including those in the United States-are already committed, I
explain the way in which this norm places important epistemic constraints on
the reasoning process by which legal decisionmakers apply laws to individual
litigants. This norm requires, among other things, that the decisionmaking
process not be arbitrary from an epistemic point of view. In other words,
nonexpert judges and juries often fail to satisfy the demands of intellectual due
process when they solicit and rely upon scientific expert testimony.
In sum, I argue that values to which legal systems are and ought to be
committed actually condemn one of the most firmly entrenched evidentiary
methods currently in place. As scientific theories continue to become more
specialized, complex, and relevant to a widening range of cases, this
incoherence between normative aspiration and actual doctrinal and institutional
procedure will increasingly threaten the legitimacy of nonexperts' legal
decisions. I conclude the Article with some brief observations about the kinds
of doctrinal and institutional reforms that might better enable legal systems to
meet the high aspirations of intellectual due process, thereby restoring, or at
least improving, the coherence of value and institutional practice.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1540 The Yale Law Journal [Vol. 107: 1535

I. THE DOCTRINAL AND INSTITuTIONAL FRAMEWORK OF

EPISTEMIC DEFERENCE

A. Evidence, Procedure, and the "Law's Epistemology"

In the course of making legal decisions, judges, juries, lawyers, and other
legal reasoners must constantly repair to factual judgments about the world.2
Indeed, if one treats judgments about what legal authorities have decided as
being within the scope of "the world," one might even conclude that every
legal decision involves some judgment about the way the world is.3 One of
the most important overall decisions legal systems must make is how, if at all,
to regulate the descriptive claims about the world-claims, that is, about how
the world is, was, or will be-that enter into the legal system. In most civil
and criminal cases, the principal doctrinal and institutional mechanisms for this
kind of regulation are rules of evidence and procedure. In virtually every area
of civil and criminal law, these are the rules that regulate judgments by legal
officials (principally judges, but also legislative officials) about which
conclusions of fact are both relevant to given legal judgments and adequately
justified. In regulating these judgments, rules of evidence and procedure
comprise what we may call the "law's epistemology"-the set of rules and

2. The advisory committee note to FederalRule of Evidence 201 makes the point nicely:
[E]very case involves the use of hundredsor thousandsof non-evidencefacts. When a witness
in an automobileaccident case says "car,"everyone,judge and jury included,furnishes,from
non-evidence sources within himself, the supplementinginformationthat the "car" is an
automobile, not a railroadcar, that it is self-propelled,probablyby an internalcombustion
engine, that it may be assumedto have four wheels with pneumaticrubbertires, and so on. The
judicial process cannotconstructevery case from scratch,like Descartescreatinga world based
on the postulateCogito, ergo sum.
FED.R. EVID.201 advisorycommittee'snote (citationomitted).
3. If assertionsabout what the law is are assertionsof fact, then there is no legal decision that does
not requiresome judgmentaboutthe way the world is, for even what are usually regardedas "purelylegal
questions,"such as questionsof law about the interpretationof statutes,are questionsabout the world. A
properanalysis of this issue concerningthe "factual"natureof the law combines the doctrinalanalysis of
rules of evidence, on the one hand, with the jurisprudentialanalysis of the concept of law, on the other.
Legal positivists characteristicallytreat statements about what the law is as statementsof fact. Jules
Coleman, for example, asserts that legal positivism is best associated with at least two theses about the
nature of law. One is the "negative"thesis that the moral acceptabilityof a norm is not a necessary
condition of its legality. The other is the "positive"thesis that "law is ultimatelya matterof social fact in
the sense that the authorityof the rule of recognition is itself a matterof social convention."Jules L.
Coleman,Rules and Social Facts, 14 HARV.J.L. & PUB. POL'Y703, 716-17 (1991). If law really is a social
fact in this sense, then knowledgeof what the law is itself is knowledgeof a type of fact-i.e., knowledge
aboutthe way (partof) the world is. This positivisttreatmentof law as fact would seem to accordwell with
the common law view that law is the kind of thing that can be "judiciallynoticed."See, e.g., Schultz v.
Tecumseh Prods., 310 F.2d 426, 433 (6th Cir. 1962) ("'The law of any State of the Union, whether
dependingupon statutesor uponjudicial opinions, is a matterof which the courtsof the United States are
bound to take judicial notice, withoutplea or proof."' (quoting Lamarv. Micov, 114 U.S. 218 (1885))).
Nonpositivistaccounts, such as RonaldDworkin's,which view law as interpretiveratherthan factual,are
not easily reconciled with the familiarevidentiarydoctrinethat law can be judicially noticed. There are,
however, ways to explain "interpretive" judgmentsas factual ones-for example, by treatinga statement
about what the law is as that interpretationof the relevantlegal materialsthat in fact makes it "the best"
it can be. See RONALDDWORKIN,LAW'S EMPIRE45-113 (1986).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1541

institutions that determine what, from a legal point of view, can be believed
with sufficient justification for the purposes of the legal system.
Ultimately, my argument is a philosophical one: When a certain condition
("the underdetermination condition")is satisfied by the aggregate testimony
of scientificwitnesses in a particularcase, thereis compellingreason to doubt
that a nonexpert legal reasoner can acquire expert informationfrom a
scientific expert in a mannerthat is both epistemicallyand legallyjustified. I
set up this philosophical analysis with a detailed and focused examination of
legal rules and institutions-chiefly those in U.S. federal law. There is an
important heuristic relation between the highly abstract philosophical analysis
presented in this Article and the rather concrete analysis of doctrine: Each
keeps the other intellectually honest. Philosophical analysis without detailed
facts is blind; recitation of detailed facts without philosophical analysis is
ignorant. Philosophical inquiry into the practices of an institution proceeds by
examining the structures of the concepts, arguments, and other intellectual
procedures that comprise those practices.4 My specific philosophical inquiry
is whether and under what conditions it is possible for scientific experts to
convey justified beliefs to nonexpert judges and juries.5 To be successful, such
inquiry must be deeply informed by at least some actual practices of the type
that are examined and explained. In this area of "legal epistemology," an
understanding of "what is actual" will materially inform the analysis of "what
is possible," just as detailed analysis of scientific theories and methods deeply
informs the philosophical analysis of science.6 Thus, an adequate philosophical
investigation of whether and how a legal system might achieve justified
epistemic deference by nonexperts to experts requires attending to the ways in
which current legal systems seek to achieve it. In this way, the detailed
doctrinal discussion in Part II will inform the philosophical inquiry that this
Article pursues throughout.
In the American federal system (to which, for the most part, I confine my
discussion of doctrine7), the rules most directly concerned with the
shepherding of experts' beliefs from their minds into those of legal

4. I discuss the distinctive philosophical "point of view" at length below. See infra Sections II.B-C.
5. The explanation of how some state of affairs or some condition (e.g., knowledge or truth) is
possible is one standard mode of philosophical explanation. For discussion of this mode, see ROBERT
NOZICK,PHILOSOPHICAL EXPLANATIONS (1981).
6. Norwood Hanson's observation about philosophy of science-"profitable philosophical discussion
of any science depends on a thorough familiarity with its history and its present state"-is no less true for
philosophical analysis of law. NoRwoOD RUSSELLHANSON,PATTERNS OF DISCOVERY:AN INQUIRYINTO
THE CONCEPTUALFOUNDATIONS OF SCIENCE3 (1958).
7. I confine my analysis to federal doctrines of evidence for two reasons. First, a great many states
have adopted rules that are very close to the federal rules and routinely look to federal court decisions
under those rules as persuasive authority. Second, the philosophical issues I raise after discussing some
basic issues in the doctrines of evidence pertain to any system seeking to bring scientific expert information
into the legal decisionmaking process. Focusing on one more-or-less unified jurisdictional system serves
to make the philosophical discussion less abstract by showing how it relates directly to an existing set of
laws.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1542 The Yale Law Journal [Vol. 107: 1535

decisionmakers are the rules of evidence that pertain to expert testimony,8

relevance,9 and judicial notice.' Close examination of two well-known U.S.
Supreme Court decisions will serve to introduce this basic rule framework and
to call attention to the nature and importance of nonexpert deference to experts
in the American legal system. One of these decisions, Daubert v. Merrell Dow
Pharmaceuticals, Inc.," interpreted key provisions of the Federal Rules of
Evidence concerning admission of expert testimony, and so its direct relevance
to my topic is no surprise. The other decision is Brown v. Board of
Education,12 a decision far better known for its momentous holding under the
Equal Protection Clause regarding governmentally mandated segregation.
Despite its more common association with equal protection doctrine, Brown
and its progeny also conspicuously illustrate the crucial importance of a court's
use of putatively scientific results in reaching and attempting to justify legal
decisions. Taken together, Daubert and Brown forcefully present the doctrinal
issues that I analyze from a philosophical point of view in the remainder of
this Article.13

B. Daubert's Philosophy of Science

Among the most important concepts in the law's epistemology are

relevance, admissibility, weight, and sufficiency of evidence. Because scientific
expert testimony is evidence, it is subject to analysis under each of these
categories-it can be relevant or not, admissible or not, sufficient or not, and
given a certain deliberative weight by those assessing it. Daubert made the
scientific reliability of scientific evidence the touchstone of its admissibility.
Before discussing Daubert itself, I briefly summarize (and oversimplify) these
four basic evidentiary concepts and the institutional procedures in which they
play a role.
Rules of evidence and procedure effect a division of decisionmaking labor
between two types of legal reasoners, each serving a distinct function, with

8. See FED R. EvID. 702-706.

9. See id. 401-403.
10. See id. 201. These are the main rules, but severalotherscontributeto the effort to effect a reliable
transferof expert informationto nonexpertlegal decisionmakers,such as the exception to the hearsayrule
for learnedtreatises.See id. 803(18) (concerningadmissionof learnedtreatises).
11. 509 U.S. 579 (1993).
12. 347 U.S. 483 (1954).
13. This importantrelationshipbetween Daubert and Brown has been largely unremarkedupon by
legal scholars, but not wholly so. GordonBeggs, for example, observes that since Daubert,
scholarshave commentedextensivelyon the expandeduse of expertevidence . . . [in] the fields
of mass tort litigation and criminal law. Less noticed, but no less dramatic,is the parallel
increase in the use of expert evidence in federal civil rights litigation. Beginning with the
landmarkcase of Brown v. Board of Education, this trend is reflected in the diverse issues
involving expert proof, often novel in nature,which have regularlyappearedin the Supreme
Court'scivil rights decisions.
Gordon J. Beggs, Novel Expert Evidence in Federal Civil Rights Litigation, 45 AM. U. L. REv. 1, 2-3
(1995) (footnotes omitted).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1543

some overlap. One function, always served by judges, is to make decisions "as
a matter of law" regarding the interpretation and application of laws, including
laws of evidence.'4 The other function is that of factfinder. This is sometimes
performed by the judge without a jury and sometimes performed by the jury
under the close supervision of the judge-supervision itself guided by rules of
evidence and procedure. Factfinders, whether judges or juries, "find" facts that
are relevant to a case-the whos, whats, whens, wheres, and whys of a
litigated dispute. These are sometimes called "evidentiary facts." Factfinders
also make judgments about whether some legal predicate like "murder" or
"breach of contract" or "negligence" applies to the (evidentiary) facts in the
way the complaining litigant claims it does.15 These are judgments about
what are sometimes called "ultimate issues," or "ultimate facts," or "mixed
questions of law and fact." For example, the judgment about whether a person
was "unreasonably searched" under the Fourth Amendment might be broken
into a law component, namely, the meaning of "unreasonable search," and a
fact component, such as where the police were standing, where the citizen was
standing, etc. Perhaps the only real importance of the division of labor into
"lawfinding" and factfinding is that appellate courts are far more deferential
to decisions made by factfinders (whether judges or juries) than they are to the
"lawfinding" decisions of lower court judges. (The reason for this difference
in deference is simple and sensible. Appellate courts are not institutionally
designed to examine testimonial evidence and other kinds of evidence firsthand
and so are far less well-situated than factfinders to make an accurate factual
judgment in the face of competing factual claims.)
Judges have special and quite significant duties vis-a-vis factfinders with
respect to proffers of evidence. It is the judge who must make threshold
decisions about which evidence proffered by a litigant is admissible and which
is not.16 Decisions about admissibility are made as a matter of law and as
such are subject to review by an appellate court (though the review of this
judgment is more lenient, more deferential, than is the review of other legal
judgments, such as the interpretation of statutory, regulatory, or constitutional
provisions or of the jurisdiction's authoritative common law doctrines). The
factfinder is permitted to consider all and only that evidence that passes this
threshold test of admissibility. This is true even when the judge herself is the
factfinder; having ruled as a matter of law that evidence is not admissible, the
judge who is also a factfinder in the case is not permitted to consider it.
Admissibility and relevance are closely related. The Federal Rules of
Evidence, for example, make the relevance of proffered evidence a necessary

14. See FED. R. EVID. 104(a).

15. For a discussion of the logic and pragmaticsof maintaininga claim at law, see Scott Brewer,
ExemplaryReasoning:Semantics,Pragmatics,and the RationalForce of Legal Argumentby Analogy, 109
HARV. L. REv. 923, 998 n.215 (1996).
16. See FED. R. EvID. 104(a).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1544 The Yale Law Journal [Vol. 107: 1535

condition of its admissibility, and make it a prima facie sufficient condition as

well."7 Like the judgment of admissibility, the judgment of relevance is made
by a judge prior to trial, when considering whether to admit proffered evidence
so that the factfinder may consider it. In the federal system, "relevant"
evidence is divided into two main categories: "logical relevance" and
"pragmatic relevance." Logically relevant evidence is evidence that has "any
tendency to make the existence of any fact that is of consequence to the
determination of the action more probable or less probable than it would be
without the evidence."1 (The modern statutory treatment of the concept of
relevance in American law obscures the traditional common law distinction
between relevance and materiality-concepts that, in their traditional distinct
meaning, will be useful in my later argument.)19 "Pragmatically" relevant
evidence is evidence whose probative value is not "substantially outweighed
by the danger of unfair prejudice, confusion of the issues, or misleading the
jury, or by considerations of undue delay, waste of time, or needless
presentation of cumulative evidence."20 A judge is permitted to exclude even
logically relevant evidence if it is not pragmatically relevant.
Once the judge decides that proffered testimonial evidence-including
testimony by experts-is admissible, the factfinder is presented with that
evidence. The factfinder listens to the screened experts and decides to what
extent to credit their testimony and what impact that testimony should have on
the case as a whole. The decision about the extent to which testimony and
other evidence should be credited is a decision about the weight of the

17. Rule 402 states in full: "All relevantevidence is admissible,except as otherwiseprovidedby the
Constitutionof the United States, by Act of Congress,by these rules, or by other rules prescribedby the
SupremeCourtpursuantto statutoryauthority.Evidence which is not relevantis not admissible."Id. 402.
18. Id. 401. Relevance is to be distinguishedfrom the burdenof proof, which is also often referred
to in probabilisticterms.The burdenof proof is the rule the factfinderuses in assessing whether,given all
the evidence that has been admitted,the complaininglitiganthas proven his assertions.
19. The modem, statutorilyenacted FederalRules of Evidence have eliminatedthe handy reference
to the distinctionbetweenmaterialityand relevancethatwas a centerpieceof the common law of evidence
(thoughthey fortunatelyhave not eliminatedthe effective legal significanceof the distinctionitself). In the
common law system, evidence was said to be "relevant"if it tendedto establishthe point for which it was
offered. It was "material"if the point for which the evidence was offered actually bore on issues in the
case. See CHRISTOPHER B. MUELLER& LAIRDC. KIRKPATRICK, EVIDENCEUNDERTHE RULES62 (2d ed.
1993). Under Rule 401, however, the requirementthat the fact proved must be "of consequence to the
determinationof the action,"FED.R. EvID.401, effectively assimilatesthe judgmentof materiality(in the
common law sense of the term) into the judgment of relevance. Thus, under the newer rules, a lawyer no
longer makes separate objections regarding the relevance (whether the proffered evidence tends to establish
the point for which it was offered) and materiality (whether the proffered evidence bears on issues in the
case). Instead, there is simply a uniform objection on grounds of irrelevance, which could really be either
an objection that the evidence is not relevant or that it is not material (in the common law senses of these
terms) or both. See John R. Schmertz, Relevancy Under 401: A Dual Concept, LITIGATION, Spring 1988,
at 12, 12. Because of this "underdeterminative" quality of the objection under the Federal Rules of
Evidence on grounds of irrelevance, it would be far more useful to maintain the old common law concepts
of materiality and relevance for both ease of reference and clarity and accuracy of evidential analysis. I use
the basic concepts of materiality and relevance (in their older, common law sense) to articulate the concept
of "rational pertinence." See infra Section V.B.
20. FED. R. EVID. 403.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1998] Scientific Expert Testimony 1545

evidence. Shouldtherebe an appealaftertrialin which the losing partyclaims

that, as a matter of law, the evidence did not support the ultimate decision
made by the judge or jury, an appellate court considers whether, as a matter
of law, the evidence was sufficient to support the verdict.
I turn now to the specific evidentiary issues in Daubert. In Daubert, the
Supreme Court considered the proper method by which federal judges are to
evaluate proffers of scientific evidence in deciding whether to admit that
evidence.21 Displacing the test that had long been used in the federal courts,
the Court held that the Federal Rules of Evidence and Federal Rules of Civil
Procedure obliged judges to ensure that proffered scientific evidence is both
"relevant" and "reliable." The complex doctrinal structure that the Court built
for the admission of scientific evidence includes Federal Rules of Evidence
104(a), 201(b),23 401,24 402,25 403,26 70227 70328 and 706(a)29a
well as Federal Rules of Civil Procedure 50(a)30 and 56.3'

21. The dispute began in the lower federal courts as an issue about the admissibility of plaintiff-
profferedexpert scientific evidence that the "morningsickness"drug Bendectincaused birthdefects. The
trial court denied the evidence's admissibilityfor failing to satisfy the Frye "generalacceptance"test for
the admissibility of novel scientific evidence and granted summary judgment for the defendant
pharmaceuticalcompany.See Daubertv. MerrellDow Pharms.,Inc., 727 F. Supp. 570, 575-76 (S.D. Cal.
1989), aff'd, 951 F.2d 1128 (9th Cir. 1991), vacated and remanded,509 U.S. 579 (1993). According to
the Frye test, a court should admit "experttestimonydeducedfrom a well-recognizedscientific principle
or discovery"only when "thething from which the deductionis made [is] sufficientlyestablishedto have
gained general acceptancein the particularfield in which it belongs."Frye v. United States, 293 F. 1013,
1014 (D.C. Cir. 1923). After the Ninth Circuitaffirmedthe lower court'sdecision in Daubert,the Supreme
Court grantedcertiorarion the question whetherthe FederalRules of Evidence, adoptedlong after Frye,
had in some way displacedthe Frye test. See Daubertv. MerrellDow Pharms.,Inc., 506 U.S. 914 (1992),
granting cert. to 951 F.2d 1128 (9th Cir. 1991).
22. The Rule reads:
Preliminaryquestionsconcerningthe qualificationof a person to be a witness, the existence of
a privilege, or the admissibilityof evidence shall be determinedby the court, subject to the
provisionsof subdivision(b) [pertainingto conditionaladmissions].In makingits determination
it is not bound by the rules of evidence except those with respectto privileges.
FED R. EVID. 104(a); see also CHRISTOPHER B. MUELLER& LAIRDC. KIRKPATRICK, EVIDENCE190 & n. 1
(1995).
23. Rule 201 reads in full: "A judicially noticed [adjudicative]fact must be one not subject to
reasonabledispute in thatit is either(1) generallyknown within the territorialjurisdictionof the trialcourt
or (2) capable of accurateand readydeterminationby resortto sources whose accuracycannot reasonably
be questioned."FED.R. EvID. 201(b).
24. See supra text accompanyingnote 18.
25. See supra note 17 and accompanyingtext.
26. See supra text accompanyingnote 20.
27. Rule 702 readsin full: "If scientific,technical,or otherspecializedknowledgewill assist the trier-
of-fact to understandthe evidence or to determinea fact in issue, a witness qualified as an expert by
knowledge, skill, experience, training, or education, may testify thereto in the form of an opinion or
otherwise."FED.R. EvID.702.
28. Rule 703 establishesthatexperts may rely on facts or data not themselvesadmissibleas evidence
"[i]f of a type reasonablyrelied upon by experts in the particularfield in formingopinions or inferences
upon the subject."Id. 703.
29. Rule 706 establishesthatthe court"mayappointany expert witnesses agreedupon by the parties,
and may appointexpertwitnesses of its own selection."Id. 706(a). Court-appointed expertsare then subject
to being called to testify by the court or by either party,and are subjectto cross examination.See id.
30. FED.R. Civ. P. 50(a) (providingfor directedverdicts).
31. Id. 56 (providingfor summaryjudgment).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1546 The Yale Law Journal [Vol. 107: 1535

Roughly, Daubert orchestrates these rules as follows. Under Rules 104(a)

and 702, a court is to assess whether the expert is qualified.32 Under Rules
702 and 703,33 as well as Rules 401, 402, and 403,34 the judge is to
determine whether the expert's methodology and principles are relevant and
"scientifically valid."35 Under Rule 706, a court is free to appoint its own
expert.36 Finally, once scientific evidence is admitted, a court may either
direct a verdict or grant summary judgment when, in the court's view, a
reasonable person could not find, on the basis of the evidence, that the position
the evidence was proffered to support was more likely than not to be true.37
Of particular concern for my purposes is how, on the Court's analysis, a
trial judge is to determine whether proffered scientific evidence is sufficiently
relevant and reliable to be presented to a jury. According to Daubert, that
determination requires the judge to admit the evidence only if it satisfies the
following criteria: It must be "ground[ed] in the methods and procedures of
science";38 must be "derived by the scientific method,"39 that is, be
"scientifically valid";40 it must be, "more than subjective belief or
unsupported speculation,"' a "body of known facts or. . . ideas inferred
from such facts or accepted as true on good grounds,"42although it need not

32. The Court'sdiscussion of the role of Rule 104(a) is somewhatunclear."Facedwith a profferof

expert scientific testimony,"said the Court,"thetrialjudge must determineat the outset, pursuantto Rule
104(a), whetherthe expert is proposingto testify to (1) scientific knowledge that (2) will assist the trier
of fact to understandor determinea fact in issue." Daubertv. MerrellDow Pharms.,Inc., 509 U.S. 579,
592 (1993) (emphasisadded)(footnotesomitted).This might suggest that"atthe outset"of every question
involving a profferof scientific evidence, a trialcourt is requiredto hold an in limine hearingunderRule
104(a). It seems that this cannot be quite right, because in many cases the methodology of proffered
evidence is well accepted in the jurisdiction,so that districtjudges will often think it not worth the cost
to the court or parties to spend time on an in limine hearing. See MargaretA. Berger, Evidentiary
Framework,in FEDERAL JUDICIAL CTR.,REFERENCE MANUAL ON SCIENTIFIC EVIDENCE 37, 50 & n.23
(1994). Assumingthe Courtis not to be takenquite at its word,Daubertdoes not give good guidanceabout
when such a hearingought to be held. Note that, when it is required,such a hearingis significantfrom an
epistemicpoint of view, becauseRule 104(a)allows the trialjudge to considerinadmissibleevidence, under
a preponderancetest, in deciding whetherto admitthe evidence. See FED. R. EVID. 104(a) ("Inmakingits
determination[the trial judge] is not bound by the rules of evidence except those with respect to
privileges."); Daubert, 509 U.S. at 592 n.10 (requiringthat judgment under Rule 104(a) "should be
established by a preponderanceof proof"). This means that a judge has much greaterfreedom to try to
inform himself in a 104(a) inquiry than in an inquiryunder other rules of evidence. See MUELLER &
KIRKPATRICK, supra note 22, at 190 n.l.
33. See Daubert, 509 U.S. at 592 ("Unlikean ordinarywitness, see Rule 701, an expert is permitted
wide latitudeto offer opinions, includingthose thatare not based on first-handknowledge or observation.
See Rules 702 and 703.").
34. See id. at 595.
35. Id.
36. See id.
37. See id. at 595-97.
38. Id. at 590. The Courttook this to be a criterionimplied by the adjective"scientific"in Rule 702.
See id.
39. Id.
40. Id. at 590-91 n.9 ("In a case involving scientific evidence, evidentiaryreliabilitywill be based
upon scientific validity.").
41. Id.
42. Id. at 590 (citing WEBSTER'S THIRD NEW INTERNATIONAL DICTIONARY OF THE ENGLISH
LANGUAGE, UNABRIDGED 1252 (1986) (defining"knowledge")).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1547

be "certain[]";43and the proffered evidence must be relevant, that is, it must

have "a valid scientific connection to the pertinent inquiry."" In sum, a judge
considering proffered scientific evidence must determine that "the reasoning
or methodology underlying the testimony is scientifically valid and...
properly can be applied to the facts in issue."45
Lest this seem too daunting a task for judges who for the most part have
no advanced scientific training, the Daubert Court saw fit to offer "general
observations," in the form of four factors (none of them either a necessary or
a sufficient condition) that a trial court should use to determine "whether a
theory or technique is scientific knowledge that will assist the trier of fact" as
required by Rule 70246 if scientific evidence consisting of or informed by that
theory or technique is to be admitted: (1) whether the scientific evidence is
testable and falsifiable;47(2) whether it has been subjected to peer review and
publication;48 (3) whether it has a high known or potential rate of error;49
and (4) whether the evidence is generally accepted in the scientific
community.50 By articulating this four-factor test for "scientific validity," the
Daubert Court effected a substantial change in the existing and longstanding
rule for the admission of scientific evidence. Under the prior rule (the so-called
Frye rule), general acceptance was the sole necessary and sufficient condition
for the admission of (novel)5' scientific evidence.52 Under Daubert's
approach, a court may, but is not required to, consider general acceptance,
along with at least three other criteria, none of which the Court treated as
either a necessary or a sufficient condition of scientific validity.53
The Daubert decision is important for my purposes in at least two ways.
First, in interpreting Rule 702's reference to "scientific knowledge," the Court
set itself the paradigmatically philosophical task of exploring the criteria of the
concept of science. A typical philosophical task is to explicate "the nature" of

43. Id.
44. Id. at 592.
45. Id. at 592-93.
46. Id. at 593.
47. See id.
48. See id.
49. See id. at 594.
50. See id.
51. Unlike the Frye Court,the DaubertCourtexpressly declined to limit its analysis to the admission
of novel scientific evidence, see id. at 592, althoughit did limit itself to Rule 702's "scientific"evidence
prong, see id. at 590 n.8.
52. See Frye v. United States, 293 F. 1013, 1014 (D.C. Cir. 1923).
53. See Daubert,509 U.S. at 591-95. The Courtassertedthat"[m]anyfactorswill bearon the inquiry,
and we do not presume to set out a definitivechecklist or test," id. at 593, and it expressly qualifiedeach
criterionlisted, see, e.g., id. (arguingthat "[o]rdinarily"a key question for determiningscientific validity
is "whetherit can be (and has been) tested");id. at 593-94 (describingpeer review and publicationas a
"pertinent"and "relevant,though not dispositive, consideration"nor the "sine qua non of admissibility");
id. at 594 (stating that "ordinarily"a court should consider the known or potential rate of error); id.
(maintainingthat "generalacceptance"can still have a bearingon the inquirysince "reliabilityassessment
does not require,althoughit does permit,explicit identificationof a relevantscientific communityand an
express determinationof a particulardegree of acceptancewithin that community"(citation omitted)).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1548 The Yale Law Journal [Vol. 107: 1535

various abstractions and the concepts used to express them, such as truth,
meaning, knowledge, justification, justice, good, bad, evil. That task typically
involves exploring the criteria that are characteristic of those concepts, and this
is precisely what the Daubert majority did in its discussion of the closely
related concepts of science and scientific knowledge. I say that the Court "set
itself' this philosophical task because engaging that task directly was not an
inevitable part of interpreting Rule 702. The Court noted, for example, that it
was interpreting Rule 702 and other "legislatively enacted Federal Rules of
Evidence as we would any statute.",54 Thus, in principle the Court could have
concluded that a particular conception of scientific knowledge-call it C-was
legally authoritative not because in the Court's view C was the correct
conception of science, but rather because in the Court's view C was the
conception of science the legislature that enacted the Rules of Evidence had
endorsed. In this way the Court would endorse a philosophical conception of
science only indirectly. Nevertheless, in order for a legal system to decide
whether certain evidence was "scientifically valid," some legal decisionmaker
along the chain of legal authority would have to undertake the philosophical
inquiry directly.
In any event, the Daubert Court appears to use a "plain meaning" method
of interpretation-hence its rather striking inquiry into the dictionary
definitions of 'knowledge'55 and 'scientific.'56 By relying so heavily on
dictionary definitions, the Court appears to have thought that its conception of
"scientific knowledge" was endorsed by the "plain meaning" of the (American)
English words 'scientific' and 'knowledge.' The decision to "look it up" might
seem incompatible with the idea that the Court was doing philosophy here at
all, but there is no necessary incompatibility between the claim that the Court's
inquiry is a philosophical one and its perhaps naive-seeming resort to the
dictionary. Investigation of ordinary meaning is by now one quite familiar
philosophical method, though, to be sure, that method can be executed with
lesser or greater skill, producing lesser or greater philosophical insights.57

54. Id. at 587.

55. I follow the philosophicalconventionof using single quotationmarksto name words. That is, I
use single quotationmarkswhen referringto words by name ratherthan using them to refer to the things
they name. Thus, whereas 'Yale' is a four-letterword, Yale is a university.I will use doublequotes to refer
to a directquotationor to referto the way in which somethingis referredto in some linguisticcommunity.
For a discussion of this convention, see WILLARDV. QUINE, METHODSOF LOGIC37-38 (1959).
56. Daubert, 509 U.S. at 590 ("The adjective 'scientific' implies a groundingin the methods and
proceduresof science. Similarly,the word 'knowledge'connotesmorethansubjectivebelief or unsupported
speculation.The term applies to any body of known facts or to any body of ideas inferredfrom such facts
or accepted as truthson good grounds."(citing WEBSTER'STHIRDNEW INTERNATIONAL DICTIONARYOF
THE ENGLISHLANGUAGE,UNABRIDGED,supra note 42, at 1252)).
57. For particularlycompelling examples of "ordinarylanguage"philosophical methods, see J.L.
AUSTIN, PHILOSOPHICAL PAPERS(J.O. Urmson& G.J. Warnockeds., 3d ed. 1979); and STANLEYCAVELL,
MUST WE MEAN WHATWE SAY? (1976). I offer additionalbrief discussionof the natureof philosophical
method below. See infra note 174.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1549

That the Court pursued the distinctly philosophical task of explicating the
criteria of the concept of science also appears from its decision to take judicial
notice of works by philosophers in offering its analysis of scientific validity.58
The Court's effort to rely on work in philosophy of science raises the question
of how, from an epistemic point of view, nonscientific expert information does
and should enter into the legal system. Although I shall not argue the point
here, there is no reason in principle to think that philosophers do not have
"technical, or other specialized knowledge" that can "assist the trier of fact to
understand the evidence or to determine a fact in issue" so that they could
usefully and relevantly serve as "witness[es] qualified as . . . expert[s] by
knowledge, skill, experience, training, or education."59And this I think is so
even though philosophy should not be counted as a "science." Although cases
in which there has been actual testimony of this sort are hard to find,60it is
much less difficult to identify instances in which judges seem to be taking
judicial notice of many different types of philosophical work.6' What criteria
a court might use to regulate admission of expert philosophical work as either
testimonial or as judicially "noticed" evidence is an important subject that
Daubert expressly declines to address.62 But Daubert's reliance on

58. Thus, for example, the Court cited the works of Karl Popper and Carl Hempel to supportthe
propositionthat testabilityand falsifiabilityare characteristicof scientific method. See Daubert, 509 U.S.
at 593 (citing C. HEMPEL,PHILOSOPHYOF NATURALSCIENCE(1966); K. POPPER,CONJECTURESAND
REFUTATIONS: THE GROwTHOF SCIENTIFIC KNOWLEDGE (5th ed. 1989)).
59. FED. R. EVID.702.
60. Hard, but not impossible. For example, two philosophers,MarthaNussbaum and John Finnis,
recently testifiedat a trial in a case raisingvariouschallenges to a state referendumthat forbadeColorado
localities from enacting civil rights protectionsspecifically for homosexuals.See David Orgon Coolidge,
Same-SexMarriage?Baehrv. Miike and the Meaningof Marriage,38 S. TEx. L. REV.1, 119 n.87 (1997).
They testified partlyas experts on ancient Greek cultureand partlyas experts on the moral acceptability
of the referendum.See JohnM. Finnis,Law,Morality,and "SexualOrientation,"69 NOTREDAME L. REV.
1049, 1056-63 (1994); MarthaC. Nussbaum,Platonic Love and ColoradoLaw: The Relevanceof Ancient
GreekNorms to ModernSexual Controversies,80 VA. L. REV.1515, 1522-24 (1994). Deep questions,of
ancient provenance,remainabout this kind of testimony.The deepest and most enduring,perhaps,is one
that comes to us in Plato's thoroughgoingcondemnationof the sophists: Can philosophy maintain its
intellectualand epistemic integritywhen it enters the public political and legal forum?
61. See, e.g., United Statesv. CharlottesvilleRedevelopment& Hous. Auth.,718 F. Supp.461, 468-69
(W.D. Va. 1989) (holding thata public housing authority'stenantassignmentplan was race-consciousand
violated the Fair Housing Act). The court noted:
While the scope of this policy advancingintegrationmust be circumscribed,thatdoes not mean
that the legal principle of integrationgoes away, that CRHA'sduty to seek integrationfades
away, or that the legal value of integrationhas no force.
To utilize an analogy from moral philosophy,when one describes the conflict of moral
principles, it is not accurateto act as if the principle which has been overriddenevaporates
withoutresidue.A principle,even after being overridden,still has some force. It still leaves, in
one philosopher'slanguage,"moraltraces."
Id. (citing Robert Nozick, Moral Complicationsand Moral Structures,13 NAT.L.F. 1 (1968)); see also
American Home Prods. Corp. v. FTC, 695 F.2d 681, 689 (3d Cir. 1982) (using Paul Grice's framework
of conversationalimplicatureto describe a false advertisingclaim).
As ProfessorLawrenceLessig has recentlyemphasized,the SupremeCourt's landmarkdecision in
Erie R.R. v. Tompkins,304 U.S. 64 (1938), relies heavily on taking judicial notice of a change in the
philosophical conception of law from a natural law to a positivist conception. See Lawrence Lessig,
UnderstandingChangedReadings: Fidelity and Theory,47 STAN. L. REV.395, 432 (1995).
62. See Daubert, 509 U.S. at 590, 592.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1550 The Yale Law Journal [Vol. 107: 1535

philosophical theories of science, albeit brief (and not particularly cogent63),

also reveals that this precise question about the admissibility or noticeability
of nonscientific expert evidence is raised at a meta-level even by Daubert's
more limited inquiry into science: It seems that the Court relied on some
expert theories in order to answer the philosophical question of what criteria
were distinctive of "science." This meta-question cannot be overlooked in a
thorough assessment of even the more limited question of the "nature" of
science, and deserves more attention than philosophers and jurists have given
it.64

Daubert is of special concern here in another way as well. It addresses

directly the question of how nonexpert judges and juries are to take account
of putatively expert scientific information, and it attempts to establish a
doctrinal and institutional system in which they can do so in an epistemically
justified manner. In so doing, Daubert raises a question that has been central
to the conscience and consciousness of American legal decisionmakers at least
since Henry Hart and Albert Sacks (and other members of the "legal process"
movement) highlighted it nearly half a century ago: the question of
institutional competence.65 Chief Justice Rehnquist's partial dissent in
Daubert (joined by Justice Stevens) immediately voiced that question in the
following terms:

I defer to no one in my confidence in federal judges; but I am at a

loss to know what is meant when it is said that the scientific status of
a theory depends on its "falsifiability," and I suspect some of them
will be, too.
I do not doubt that Rule 702 confides to the judge some
gatekeeping responsibility in deciding questions of the admissibility
of proffered expert testimony. But I do not think it imposes on them
either the obligation or the authority to become amateur scientists in
order to perform that role.66

Judge Kozinski, who wrote the opinion upon remand from the Supreme
Court in Daubert,67 put this question of institutional competence even more

63. The Court shows no awareness of post-positivistcritiques of Popper's falsifiability thesis, for
example. One of the most influentialof these critiques is W.V. QUINE, Two Dogmas of Empiricism,in
FROM A LOGICALPOINTOF VIEW: 9 LOGICO-PHILOSOPHICAL ESSAYS20 (2d ed. 1961).
64. Anthony Kenny may be correct that "[t]he difficulty of deciding in the courtroom whether
something is a science is not a difficulty which could be solved by admitting philosophers into the
courtroomas higherorderexperts,"AnthonyKenny,The Expertin Court,in THE IVORYTOWER:ESSAYS
IN PHILOSOPHY AND PUBLICPOLICY51 (1985), but surelythe admissionof such testimonycould help solve
the problem. Kenny also contends that "[t]he decision whethersomething is or is not a science is . .. a
matterfor philosophy,"and that, on the test Kenny offers for science, "[pihilosophyitself is clearly ...
no science [since it is] neitherconsistent, methodological,cumulative,or predictive."Id.
65. See HENRY M. HART & ALBERTM. SACKS, THE LEGAL PROCESS:BASIC PROBLEMSIN THE
MAKINGAND APPLICATION OF LAW (William N. Eskridge& Philip P. Frickeyeds., 1994).
66. Daubert, 509 U.S. at 600-01 (Rehnquist,C.J., concurringin part and dissentingin part).
67. See Daubert v. Merrell Dow Pharms.,Inc., 43 F.3d 1311 (9th Cir. 1995). Judge Kozinski also

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1551

pointedly:

The first prong of Daubert puts federal judges in an uncomfortable

position. The question of admissibility only arises if it is first
established that the individuals whose testimony is being proffered are
experts in a particular scientific field.... Yet something doesn't
become "scientific knowledge" just because it's uttered by a scientist;
nor can an expert's self-serving assertion that his conclusions were
"derived by scientific method" be deemed conclusive .... As we
read the Supreme Court's teaching in Daubert, therefore, though we
are largely untrained in science and certainly no match for any of the
witnesses whose testimony we are reviewing, it is our responsibility
to determine whether those experts' proposed testimony amounts to
"scientific knowledge," constitutes "good science," and was "derived
by the scientific method."68

The question, then, is this: How could a scientifically untrained judge be

sufficiently epistemically competent to perform the gatekeeping task imposed

wrote the originalDaubert opinion to which the SupremeCourtgrantedcertiorari.See Daubertv. Merrell

Dow Pharms.Inc., 951 F.2d 1128 (9th Cir. 1991), vacated and remanded,509 U.S. 579 (1993). In that
opinion, he affirmedthe districtcourt'sgrantof summaryjudgment,applyingthe Frye test to hold thatthe
profferedscientific evidence was not admissible.See supra note 21. On remandfrom the SupremeCourt,
Judge Kozinski's panel decided to apply the newly minted test withoutfirst remandingto the district
court-offering as a justificationfor this proceduralarrogationthe contentionthat it was permittedby "the
peculiar circumstancesof the case," "the interestsof justice," and "judicialeconomy."Daubert, 43 F.3d
at 1314-15.
Given the newness of the test, and the possibility that the originalplaintiffsmight have been able to
develop new evidence or repackageexisting evidence pursuantto it, the wiser course would have been to
remandto the trial court. Apartfrom the claimed equities, Judge Kozinski's substantiveargumentfor his
decision to decide the case withoutremandingto the trialcourtwas as follows. The Dauberttest, according
to Judge Kozinski, has "two prongs":First, the plaintiffs must show that the profferedscientific evidence
really is "derivedby the scientific method";second, they must show that the evidence would "assist the
trier of fact." Id. at 1320. Judge Kozinski conceded that the plaintiffs might have been able to "submit
additionalproof that the scientific testimonythey proffer[ed]was 'derivedby the scientific method,"'but
he also concluded that they could not possibly show that their evidence would "assist the trier of fact,"
mainly because the evidence they could adduce would not meet the burdenof proof for causationunder
Californiatort law. Id. at 1320-22.
This analysis is unconvincing.JudgeKozinski'sclaim thathe could decide whetherthe evidence was
relevant without seeing whether it was sufficiently scientifically valid seems to overlook the likely
intellectual synergy between the "prongs"of the Daubert analysis. (I assume, arguendo, that Judge
Kozinski's simplificationof the Daubert rule is accurate-it seems actually to be an oversimplification.)
If plaintiffs could, under the new and fairly detailed guidance of Daubert, show that their evidence was
indeed "derivedby the scientific method,"such a demonstrationmight very well have helped to show that
the evidence could also "assist the trier of fact." At least, such interactionbetween the two showings is
clearly possible and not unlikely in many instances. Moreover,a careful readingof the SupremeCourt's
Daubert opinion shows that the Courtitself clearly saw a close enough connectionbetween the "scientific
validity" criterionof Rule 702 and the "helpfulness"criterionas to suggest that it is a mistake to try to
assess them independently.See Daubert, 509 U.S. at 591-92.
68. Daubert, 43 F.3d at 1315-16. Kozinski here echoes the concerns not only of Chief Justice
Rehnquistand Justice Stevens, but also those of a great many commentators.See, e.g., Paul S. Milich,
ControversialScience in the Courtroom:Daubertand the Law's Hubris, 43 EMORYL.J. 913, 919 (1994)
("Scientists who have spent . . . their professional lives wrestling with the . . . mysteries of their disciplines
must be amazed at the law's hubrisin thinkingthat nonscientistjudges can . . . ultimatelydecide who has
the better. . . argument.").

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1552 The Yale Law Journal [Vol. 107: 1535

upon it by Daubert's reading of the Federal Rules of Evidence? Moreover,

assuming the judge admits the evidence, how could a scientifically untrained
trier of fact, whether judge or jury, be sufficiently epistemically competent to
assess competing putatively scientific claims by competing expert witnesses
when, ex hypothesi, that factfinder does not have the requisite expertise to
judge the evidence itself?69
Although the Daubert majority expressed its "confiden[ce] that federal
judges possess the capacity to undertake" the analysis of proffered scientific
evidence described above,70 nowhere in its opinion does it provide any
support for that judgment. The opinion indirectly addresses the question when
it considers whether its elimination of "general acceptance" as a necessary
condition for the admission of scientific evidence "will result in a 'free-for-all'
in which befuddled juries are confounded by absurd and irrational
pseudoscientific assertions."7' Such a view, the Court explains, is

overly pessimistic about the capabilities of the jury and of the

adversary system generally. Vigorous cross-examination, presentation
of contrary evidence, and careful instruction on the burden of proof
are the traditional and appropriate means of attacking shaky but
admissible evidence.... Additionally, in the event the trial court
concludes that the scintilla of evidence presented supporting a position
is insufficient to allow a reasonable juror to conclude that the position
more likely than not is true, the court remains free to direct a
judgment ... and likewise to grant summary judgment .... These
conventional devices, rather than wholesale exclusion under an
uncompromising "general acceptance" test, are the appropriate
safeguards where the basis of scientific testimony meets the standards
of Rule 702.72

These remarks, however, wholly fail to show how it is possible for a judge
to assess scientific validity (or for a factfinder to "adjudicate" competing
claims to scientific truth) in an epistemically justified manner. The Court
asserts that "vigorous cross-examination, presentation of contrary evidence, and
careful instruction on the burden of proof' are "appropriate" means for
"attacking shaky but admissible evidence." But if a judge or a jury does not
have the requisite scientific training, how can that judge or jury make a

69. I say "ex hypothesi"here because if the factfinderdid have thatexpertise,the evidence would not
be necessary and should thus probably be excluded under Rule 403. See FED. R. EVID. 403 (making
relevantevidence excludable "if its probativevalue is substantiallyoutweighed by ... considerationsof
undue delay, waste of time, or needless presentationof cumulativeevidence").Rule 702 also might afford
a groundfor excluding the evidence; for example, assumingthe law of thermodynamicswere relevantto
a case, and the trierof fact were a physicist, it would not "assistthe trierof fact"to have expert testimony
restatingthat law. See id. 702.
70. Daubert, 509 U.S. at 593.
71. Id. at 595-96.
72. Id. at 596.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1553

warranted choice between competing "vigorously cross-examined" claims by

putative experts in, say, medicine, mathematics, chemistry, or biology?73 Of
what use is "careful instruction on the burden of proof' to a jury that cannot
understand the formal methodology used to reach a scientific conclusion? How
can a judge who is, in Judge Kozinski's words, "largely untrained in science
and certainly no match for any of the witnesses whose testimony [he is]
reviewing," justifiably direct a verdict or grant summary judgment against a
party in the face of contrary contentions by that party's expert? I explore these
precise questions in greater detail below when I undertake a more robustly
philosophical inquiry into epistemic deference. Here I want only to set the
stage for that argument by showing precisely how Daubert itself raises these
issues-and, in the next section, how Brown v. Board of Education74 and its
progeny raise them as well.

C. Brown's Empirical Psychology

Although the main Brown opinion does not expressly address the use by
courts of putatively scientific expert evidence (and although Brown was
decided long before the Federal Rules of Evidence were adopted in 197575),
the use of such information in Brown and subsequent cases decided in its wake
illustrates in a pointed way both the hazards and the vast importance of this
reasoning process.
The facts and basic legal conclusions of Brown are too well-known to
warrant recital here. What I shall focus on instead is the rather striking extent
to which, at least on the argumentative surface of the main Brown opinion, the
Court rested its holding on contested judgments of empirical social science. In
so doing, Brown served as a remarkable culmination of the legal realist project
of taming abstract legal propositions with the whip of social science-a
process that began in the modern Supreme Court with the Court's acceptance
of the "Brandeis brief' in Muller v. Oregon.76

73. Juristsconcerned with the rules and institutionsof expert evidence have long been wary of the
adversary system's capacity to produce truths. Professor John Langbein's skeptical inquiry is a fair
representationof this concern:
Wigmore's celebrated panegyric-that cross-examinationis "the greatest legal engine ever
invented for the discovery of truth"-is nothing more than an article of faith .... Judge
Frankelexplains why: "The litigator'sdevices, let us be clear, have utility in testing dishonest
witnesses, ferretingout falsehoods, and thus exposing the truth.But to a considerabledegree
these devices are like other potent weapons, equally lethal for heroes and villains." . . . In the
hands of many of its practitioners,cross-examinationis not only frequentlytruth-defeatingor
ineffectual, it is also tedious, repetitive,time-wasting,and insulting.
John H. Langbein, The GermanAdvantagein Civil Procedure,52 U. CHI.L. REV.823, 833 n.31 (1985)
(citations omitted).
74. 347 U.S. 483 (1954).
75. See An Act To EstablishRules of Evidence for CertainCourtsand Proceedings,Pub. L. No. 93-
595, ? 3, 88 Stat. 1926, 1926 (1975).
76. 208 U.S. 412 (1907). For a partial reprintand discussion of the brief, see JOHNMONAHAN &
LAURENSWALKER,SOCIALSCIENCEIN THELAW 187-90 (4th ed. 1998).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1554 The Yale Law Journal [Vol. 107: 1535

Two empirical judgments undergirded the Supreme Court's decision in

Brown, both of them framed by the Court as judgments that there had been
deeply significant changes in society and in the state of social-scientific
knowledge between the time of Plessy v. Ferguson77 and the time of Brown
itself. One was that a significant change had occurred in the role of public
education in American life.78 The second, more celebrated judgment, is the
one the Court made, relying on the social-scientific authority of sociologists
(such as Gunnar Myrdal) and psychologists (such as Kenneth Clark), that
segregated schools inflict psychological harm on black schoolchildren in such
a way as to make segregated schools for black children "inherently unequal"
to segregated schools for white children.79One striking feature of the Brown
opinion is that Chief Justice Warren nowhere asserts that Plessy was wrong as
a matter of law in its interpretation of the Constitution.80 The only error
charged against the opinion is one regarding an empirical psychological claim:

Whatever may have been the extent of psychological knowledge at the

time of Plessy v. Ferguson, this finding is amply supported by modern
authority. Any language in Plessy v. Ferguson contrary to thisfinding
is rejected.
We conclude that in the field of public education the doctrine of
"separate but equal" has no place. Separate educational facilities are
inherently unequal. Therefore, we hold that the plaintiffs . . . are, by
reason of the segregation complained of, deprived of the equal
protection of the laws guaranteed by the Fourteenth Amendment.8'

77. 163 U.S. 537 (1896).

78. The Courtnoted:
[W]e cannot turn the clock back to 1868 when the [Fourteenth]Amendmentwas adopted,or
even to 1896 when Plessy v. Ferguson was written.We must considerpublic educationin the
light of its full developmentand its presentplace in Americanlife throughoutthe Nation....
Today, education is perhaps the most important function of state and local
governments.... In these days, it is doubtfulthat any child may reasonablybe expected to
succeed in life if he is denied the opportunityof an education.
Brown, 347 U.S. at 492-93.
79. See id. at 494 n.1I.
80. This is a point ably made in PAUL L. ROSEN,THE SUPREMECOURTAND SOCIALSCIENCE163-64
(1972).
81. Brown, 347 U.S. at 494-95 (emphasis added) (footnote omitted). The famous footnote 11 of
Brown,in which the Courtcited to social scientific literaturein supportof its conclusions,readsas follows:
K. B. Clark,Effect of Prejudiceand Discriminationon PersonalityDevelopment(Midcentury
White House Conferenceon Childrenand Youth, 1950); Witmerand Kotinsky,Personalityin
the Making (1952), c. VI; Deutscher and Chein, The Psychological Effects of Enforced
Segregation:A Survey of Social Science Opinion,26 J. Psychol. 259 (1948); Chein, What are
the PsychologicalEffects of SegregationUnderConditionsof EqualFacilities?,3 Int.J. Opinion
and Attitude Res. 229 (1949); Brameld, EducationalCosts, in Discriminationand National
Welfare(Maciver, ed.,1949), 44-48; Frazier,The Negro in the United States (1949), 674-681.
And see generally Myrdal,An AmericanDilemma (1944).
Id. at 494 n.11. The plaintiffs in Brown strongly encouragedthe Courtto lean heavily on social science
(when the law is against you, as the saying goes, argue the facts). They filed, for example, a detailed
appendixto their brief, titled TheEffectsof Segregationand the Consequencesof Desegregation:A Social
Science Statement.See Appendixto Appellants'Briefs, Brown (No. 8). Similarly,in argumentbefore the
Court,ThurgoodMarshallassertedthat the social science witnesses who had testified below "standin the

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1555

Because I am interested here in the question of whether judges and

juries can use expert information in an epistemically justified way, it is
useful to pause to see how the expert information on which the Brown
Court so heavily relied actually entered into and proceeded through the
lengthy multi-court decisionmaking process. First, as to the judgment about
the widespread impact of education in modem society, it seems that the
Supreme Court took judicial notice.82 Although there is likely to have
been some discussion of that issue by the experts who testified in the trial
courts and some record of such testimony before the Supreme Court, the
proposition is asserted with bold strokes, wholly without citation or even
mention of any kind of study, finding, conclusion, or brief. In this way,
this conclusion is noticeably different from the other central judgment in
the case, that of the psychological harm. Although it is clear from the
Brown opinion that the Court was claiming to defer epistemically,
regarding that second judgment, to the judgments of "expert" social
scientists (that is, claiming to take some propositions to be true because
the social scientists said they were true), it is somewhat unclear from the
Court's opinion how exactly it understood the procedural and evidentiary
devices it was using to bring this expert information into its decision. It
is unclear, for example, whether the Court understood itself to be deferring
directly to the social-scientific expertise of Kenneth Clark and others,83
or whether instead it was only indirectly deferring to that expertise by
deferring to factual findings made in some of the decisions below. Clark
had testified for the plaintiffs in Briggs v. Elliot,84 one of the companion
cases that went up to the Court in Brown, and there was similar
testimony in other companion cases, including cases from Kansas,85

recordunchallengedas experts in their field, and I think we have arrivedat the stage where the courts do
give credence to the testimony of people who are experts in their fields." ARGUMENT:THE ORAL
ARGUMENTBEFORETHESUPREMECOURTIN BROWNv. BOARDOF EDUCATION OF TOPEKA, 1952-55, at 36,
37 (Leon Friedmaned., 1969) [hereinafterBROWNORAL ARGUMENT].
82. Paul Rosen characterizesit thus. See ROSEN, supra note 80, at 162. In the terms established by
the currentFederalRules of Evidence, this would be judicial notice of a "legislativefact,"and thus would
not be in any way regulatedby the rules. See FEDR. EVID.201(a) advisorycommittee'snote.
83. See supra note 81 and accompanyingtext.
84. 98 F. Supp. 529 (E.D.S.C. 1951), vacated, 342 U.S. 350 (1952). Briggs was the South Carolina
case in which the court upheld the constitutionalityof segregationand expressly refusedto overturnit on
the basis of testimonyby putativeeducationaland sociological experts.Like the othercompanioncases in
Brown, this court's treatmentof the expert evidence stands in instructivecontrastto the treatmentby the
Court in the Brown opinion:
We conclude, therefore,that if equal facilities are offered, segregationof the races in the public
schools as prescribedby the Constitutionand laws of South Carolinais not of itself violative
of the FourteenthAmendment.We think that this conclusion is supportedby overwhelming
authoritywhich we are not at liberty to disregardon the basis of theories advancedby a few
educatorsand sociologists. Even if we felt at libertyto disregardotherauthorities,we may not
ignore the unreverseddecisions of the SupremeCourtof the United States which are squarely
in point and conclusive of the question before us.
Id. at 536-37.
85. The Kansascase was Brownv. Board of Education,98 F. Supp. 797 (D. Kan. 1951). In thatcase,

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1556 The Yale Law Journal [Vol. 107: 1535

Virginia,86 and Delaware.87 On the basis of this procedural posture one

might conclude that the Brown Court had reviewed some of the empirical
psychological testimony under the deferential "clearly erroneous" standard that
appellate courts accord to trial courts regarding questions of fact.88 The
Court's opinion goes some way toward suggesting this by quoting a finding in
one trial court case about the psychological hardship on black school
children,89and then declaring that "this finding is amply supported by modern

the court admittedexpert testimony regardingthe psychological harm of public school segregation but
upheld the constitutionalityof segregation.See infra note 89 (noting the SupremeCourt'squotationof the
districtcourt's discussion of this testimony).
86. The Virginiacase was Davis v. CountySchool Board, 103 F. Supp. 337 (E.D. Va. 1952). That
court admittedconflictingexpert testimonyon the questionof the harmof segregationand decided it was
a toss-up on that evidence (so that, given the burdenof persuasion,the plaintiffs failed to establish the
proposition).The opinion provides a nice contrastto the use of social science evidence by the Supreme
Court in Brown, so I quote it at some length:
Eminenteducators,anthropologists,psychologistsand psychiatristsappearedfor the plaintiffs,
unanimously expressed dispraise of segregation in schools, and unequivocally testified the
opinion that such separation distorted the child's natural attitude, throttled his mental
development, especially the adolescent, and immeasurably abridged his educational
opportunities.For the defendants,equally distinguishedand qualifiededucationistsand leaders
in the other fields emphatically vouched the view that, given equivalent physical facilities,
offerings and instruction,the Negro would receive in a separateschool the same educational
opportunityas he would obtain in the classroomand on the campus of a mixed school. Each
witness offered cogent and appealinggroundsfor his conclusion.
On this fact issue the Court cannot say that the plaintiffs' evidence over-balancesthe
defendants'.But on the same presentationby the plaintiffsas just recited, Federalcourts have
rejected the proposition,in respect to elementaryand junior high schools, that the required
separationof the races is in law offensive to the Nationalstatutesand constitution.They have
refused to decree that segregationbe abolishedincontinently.We accept these decisions as apt
and able precedent.Indeed we might groundour conclusion on their opinions alone. But the
facts proved in our case, almost without division and perhaps peculiar here, so potently
demonstratewhy nullificationof the cited sections of the statutesand constitutionof Virginia
is not warranted,that they should speak our conclusion.
Id. at 338-39 (citation omitted).
87. See Belton v. Gebhart,87 A.2d 862 (Del. Ch. 1952); see infra note 89 (providingthe Supreme
Court'squotationfrom the Delawarecourt regardingthat court's assessmentof the expert testimony).
88. Note thatthe clearlyerroneousstandardof FED.R. Civ. P. 52(a) was in place at the time of Brown
and routinelyused in federal courts.
89. The Brown Courtreasoned:
To separate them from others of similar age and qualificationssolely because of their race
generatesa feeling of inferiorityas to their statusin the communitythatmay affect their hearts
and minds in a way unlikely ever to be undone. The effect of this separation on their
educationalopportunitieswas well stated by a finding in the Kansas case by a court which
neverthelessfelt compelled to rule againstthe Negro plaintiffs ....
Brown v. Board of Educ., 347 U.S. 483, 494 (1954). The Courtquoted,withoutcitation, the Kansascourt
as follows:
"Segregationof white and colored childrenin public schools has a detrimentaleffect upon tIhe
colored children.The impact is greaterwhen it has the sanction of the law; for the policy of
separatingthe races is usually interpretedas denotingthe inferiorityof the negrogroup.A sense
of inferiorityaffects the motivationof a child to learn. Segregationwith the sanction of law,
therefore,has a tendencyto [retard]the educationaland mentaldevelopmentof negro children
and to deprivethem of some of the benefitsthey would receive in a racial[ly]integratedschool
system."
Id. In addition, the Supreme Court added the following footnote: "A similar finding was made in the
Delaware case: 'I conclude from the testimonythat in our Delaware society, State-imposedsegregationin
education itself results in the Negro children, as a class, receiving educationalopportunitieswhich are
substantiallyinferior to those available to white children otherwise similarly situated."'Id. at 494 n.10

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1557

authority."90Perhaps the Court's heavy and highlighted reliance on putative

social-scientific facts was part of a realpolitik rhetorical strategy. The careful
student of Brown should not overlook the possibility that the Court went out
of its way to quote the "findings" in the Kansas and Delaware cases to build
some political cover-perhaps hoping that by appearing to defer to factual
findings by trial courts, the Court might seem less like a roving activist
tribunal searching for "facts" that suited its results. Moreover, this might well
have been a wise judgment. Writing in the third century in which science
enjoyed its ascendancy over religion as the dominant cultural authority, the
Court might reasonably have sought to invoke social-scientific expertise to
provide cultural authority for its profoundly controversial decision.
But even if the Court did in part seek the appearance of appellate modesty
in the face of scientific facts "found" by trial courts, it also did not stand-or
hide-behind the procedural device of a deferential standardof review. Despite
its reference to the findings below, the Court also clearly asserted in its own
epistemic voice that those findings were "amply supported by modern
authority."91Thus, the Court, at least to some extent, took "judicial notice"
of what would today be called "legislative facts."92By whichever procedural
or evidentiary devices social-scientific expert information entered the Court's
decision in Brown, the substantial extent to which the Court claimed to rely on
this information was of great moment in the history of the law's epistemology.
How well did the Court do in making an accurate psychological judgment
based on the expert information it had before it? The conclusion that
segregation caused psychological harm appears to have been shared among a
great many social scientists,93 and the finding seems to have withstood the
test of time.94 On the other hand, the studies on which the Court relied had
serious limitations. Surely one of the most significant of these limitations is to

(quoting Belton, 87 A.2d at 865).

90. Id. at 494.
91. Id.
92. See FED.R. EvID.201(a) advisorycommittee'snote. In colloquy with ThurgoodMarshallduring
the first oral argumentin Brown, Justice Frankfurterasked, "Canwe not take judicial notice of writings
by people who competentlydeal with these problems?Can I not take judicial notice of Myrdal's book
without having him called as a witness?"BROWN ORALARGUMENT, supra note 81, at 63. After Marshall
batted this softball question with cautiousenthusiasm,Frankfurterwent on to assertthat "in these matters
this Courttakesjudicial notice of accreditedwritings,and does not have to call the writersas witnesses."
Id. He added, "How to inform the judicial mind, you know, is one of the most complicatedproblems.It
is betterto have witnesses, but I did not know that we could not readthe worksof competentwriters."Id.;
see also ROSEN,supra note 80, at 141.
93. The Deutscherand Chein study,which the Courtcited thirdin footnote 11, see supra note 81, was
a survey of more than 800 social scientists regardingtheir views about the psychological effects of
segregation. See Max Deutscher & Isidor Chein, The Psychological Effects of EnforcedSegregation:A
Survey of Social Science Opinion, 26 J. PSYCHOL.259 (1948). Of that number,517 responded,90% of
whom agreedthatsegregationhad a harmfuleffect on the segregatedgroup.See id.; see also ROSEN,supra
note 80, at 188. This kind of unanimitymight even have allowed this factualclaim to be admittedas expert
testimony underthe Frye rule-indeed, perhapsin some of the Browntrial courts it was.
94. See ROSEN,supra note 80, at 194-95 (listing studies).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1558 The Yale Law Journal [Vol. 107: 1535

be found in the celebrated doll study, which the Brown opinion lists first in its
citation of "ample authorities" in footnote eleven.95 In that study, Kenneth
and Mamie Clark sought to measure the self-conception of two groups of black
schoolchildren, one attending integrated schools in the North, the other
attending segregated schools in the South.96 The Clarks asked the children
whether they liked to play with a brown doll or a white doll, which of the
dolls was the "nice doll," which was the doll that "looks bad," and which was
the doll that had a "nice color."97 What survives in the "popular"
understanding of this study among people familiar with the Brown opinion is
that this study showed that black children in segregated schools felt worse
about themselves, as measured by their responses to these questions, than did
the black children in integrated schools in the North. Some such finding, of
course, would be the principal relevance, if not the only relevance, of such a
study-and Kenneth Clark was the leading testimonial expert in many of the
cases leading to Brown. The actual result of the Clarks' research, however, was
quite different. Comparing the responses of black schoolchildren in segregated
southern schools to those of black schoolchildren in integrated northern
schools, the Clarks' study candidly stated:

A significantly higher percentage (71) of the northern children,

compared to the southern children (49), think that the brown doll
looks bad .... Also a slightly higher percent of the southern children
think that the brown doll has a "nice color" [the numbers here are 40
and 37, respectively] ....
The southern children, . . . in spite of their equal favorableness
toward the white doll, are significantly less likely to reject the brown
doll (evaluate it negatively), as compared to the strong tendency for
the majority of the northern children to do so.98

Both Thurgood Marshall and Clark himself were aware of the problem that
this expert evidence posed for the legal argument about the psychological
harms of segregation.99 These results suggested not only that black
schoolchildren might not be psychically harmed by segregated schooling, but
that they might in fact be harmed by integrated schooling. Clark's own
explanation of this finding vis-a-vis his larger claim about the harm of
segregation was that it would be a mistake to conclude from his findings that

95. See supra note 81.

96. See KennethB. Clark& Mamie P. Clark,Racial Identificationand Preferencein Negro Children,
in READINGSIN SOCIALPSYCHOLOGY602 (EleanorE. Maccoby et al. eds., 3d ed. 1958).
97. Id.
98. Id. at 611 (emphasisadded).
99. See RICHARDKLUGER,SIMPLEJUSTICE:THE HISTORYOF BROWN V. BOARD OF EDUCATION AND
BLACK AMERICA'SSTRUGGLEFOR EQUALITY356 (1975) ("Marshallhad weighed the risks of Clark's
findings and decided that on balance they demonstratedinjuryto segregatedNegro youngsters. 'I wanted
this kind of evidence on the record,' he had said.").

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1559

southern black schoolchildren in segregated schools suffered less psychic harm

than their northern integrated counterparts.1? Such a conclusion, he
maintained,

would seem to be not only superficial but incorrect. The apparent

emotional stability of the southern Negro child may be indicative only
of the fact that through rigid racial segregation and isolation he has
accepted as normal the fact of his inferior social status. Such an
acceptance is not symptomatic of a healthy personality.101

If not flatly question-begging, this conclusion is far too ad hoc to provide

solid ground for the contention it was adduced in the Brown litigation to
support, namely the claim that segregation in schools caused psychological
harm to black schoolchildren.102 In effect, from an epistemic point of view,
instead of strongly supporting the claim that segregated schooling caused
psychological harm to black schoolchildren, the Clarks' study actually provided
reason (at least prima facie reason) to doubt that conclusion. Although the
study seems particularly vulnerable in this way, the difficulty of showing the
independence of causes besets virtually all the empirical data proffered in
Brown. Moreover, it is a problem endemic to most social science findings
where controlled and representative experiments are rarely possible.
Neither the weakness of the particular Clark study and testimony relied on
so heavily in the Brown litigation, nor the general limitations of social-
scientific methodology by itself, warrant the conclusion that the Court was
wrong to use social-scientific data in Brown, or that other courts should not
rely on such data in general. It is possible to offer a plausible defense of the
precise use to which the Court put these studies in Brown, and some writers
have done so. Paul Rosen, for example, defends the Brown Court's use of
these data on the following ground. He acknowledges that social science is
significantly more limited than natural science and that the studies relied on in
Brown had substantial methodological limitations. Specifically, they failed to
identify school segregation as an independent cause of whatever psychic harms
resulted from segregation.103But he also observes that the studies cited in

100. See KENNETH B. CLARK,PREJUDICEAND YOUR CHILD45 (2d ed. 1963).

101. Id.
102. Moreover, there seem to be some design flaws in the study that one might readily identify
(though these flaws are perhaps more readily noticeable with the benefit of more than 40 years of
hindsight).One mightwell expect-and perhapsone mightreasonablyhave expectedeven then-that black
children in integratedschools in the North would atfirst be deeply traumatizedby being throwntogether
with white childrenin a generalculturalatmosphereof discrimination.But one mightalso expect (and have
reasonablyexpected then) that over time the effect would be to reducethe psychic discomfortof being in
an integratedsetting,therebyallowing integrationto give greaterpsychic sustenanceto blackschoolchildren
in the long run. The questionthis possibility raises for the design of the Clarkstudy is, over what span of
time must the childrenbe studiedbeforeone can reacha warrantedresultaboutthe specific harmsallegedly
arising from segregation?(Thanksto Henry Steinerfor helpful discussion on this point.)
103. See ROSEN, supra note 80, at 138-39.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1560 The Yale Law Journal [Vol. 107: 1535

Brown were intellectually honest enough to wear their methodological limits

on their sleeves and that the Court's finding of psychological harm was
consistent both with what the studies modestly suggested and with what the
overwhelming majority of social scientists believed about the psychic harms
of segregation. Warren's opinion did not, for example, contend that segregated
schooling was the sole cause of psychological harm.104Rosen further argues
that unless one believes that no social science is to be accepted unless it rises
to the level of confidence one can achieve with research in the natural
sciences, one should not reject the use of these studies, and maintains that the
few social scientists who actively opposed the conclusions of these studies
were themselves strong, ideologically biased segregationists.105 (It is fair
game, of course, to ask whether Clark and other social scientists were also
ideologically biased in their analyses.)106
This modest defense of the Brown Court's use of expert evidence is in my
view quite plausible. But there is nevertheless one very important way in
which the Court's use of this evidence, in Brown itself and in subsequent
cases, suggests a serious overstatement of the limited empirical result that
Clark and others sought to confirm. It is just this implied overstatement that
raises the issue of the justifiability of epistemic deference that is of central
concern in this Article.
What I have in mind here is the Court's assertion in Brown-on the basis
of sharply and openly limited empirical data-that separation is "inherently"
unequal.107 How could the Court legitimately conclude that separation is
inherently unequal when the trumpeted basis for its judgment was contingent
empirical evidence that was, and should have been recognized to have been,
subject to further confirmation or disconfirmation? Perhaps this problem is in
part terminological. Perhaps the Court should have said something like,
"according to the best data currently available," instead of "inherently," or
perhaps that is what the Court's assertion should be taken to mean. But even
such a change would not eliminate the disproportion between the empirically
modest basis for the legal conclusion in Brown, and the powerful, far-reaching
conclusion that the Court sought to make virtually unquestionable.
That the Court did, in effect, make it unquestionable is an important part
of my concern here. Given that Brown seemed to have rested its opinion so
firmly on contingently verifiable psychological facts,108 lower courts, and
members of the polity more generally, might well have thought it legitimate
to "help" the Court refine and, if necessary, correct the scientific-factual

104. See Brown v. Board of Educ., 347 U.S. 483, 492-93 (1954).
105. See ROSEN, supra note 80, at 141-43.
106. See id. at 182-96.
107. Brown, 347 U.S. at 494-95.
108. See id. at 494 ("Whatevermay have been the extent of psychological knowledge at the time of
Plessy v. Ferguson, this finding is amply supportedby modem authority.").

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1561

judgments it had made so central a part of American constitutional life. Indeed,

nearly ten years after Brown, a federal district court judge in Georgia accepted
exactly that implied invitation.
In Stell v. Savannah-Chatham County Board of Education,109 the district
court denied an injunction to prevent a local school board from operating a
segregated school system. The court took evidence and explicitly held, in the
face of Brown, that the best available evidence established that segregated
schooling caused no injury to white or black children, and further that forced
integration did cause harm to both groups. The court reasoned that Brown
rested on a scientific factual finding of psychological harm"0 and that the
doctrine of stare decisis did not bind nonparties to a case's factual findings. As
a result, later courts were free to revisit the issue.111Moreover, the court
concluded that the preponderance of the evidence-including Clark's own
study-showed by "uncontroverted testimony on scientific issues given as the
unanimous opinions of conceded authorities open to cross-examination on the
witness stand" that there was no psychological harm of the sort Brown had
claimed."2 Accordingly, the judge "accept[ed] the evidence given in the

109. 220 F. Supp. 667 (S.D. Ga. 1963), rev'd, 333 F.2d 55 (5th Cir. 1964), cert. denied, 379 U.S. 933
(1964).
110. After noting the Brown opinion's quotationfrom the Kansascourt, the districtjudge asserted:
These are facts, not law. To make these findings the Kansas District Judge considered
evidence-not cases. Whether Negroes in Kansas believed that separate schooling denoted
inferiority, whether a sense of inferiority affected their motivation to learn and whether
motivation to learn was increased or diminished by segregation was a question requiring
evidence for decision. That was as much a subjectof scientific inquiryas the brakingdistance
requiredto stop a two-ton truckmoving at ten miles an hour on dry concrete.
Again, the Supreme Court quoted the record that Negro children in Delaware were
"receivingeducationalopportunitieswhich are substantiallyinferiorto those availableto white
children otherwise similarly situated,"a statementwhich could only be of factual ratherthan
legal significance.
The SupremeCourtput at rest any residualquestionon the natureof its inquirywhen it
indicated its reliance on scientific information:"Whatevermay have been the extent of
psychological knowledge at the time of Plessy v. Ferguson,this findingis amply supportedby
modem authority."The teachings of psychology in 1896, in 1954, or in 1963 are inquiries
requiringevidence in the same sense as repeateddeterminationsof "seaworthiness."Actually,
the non-legal authorityto which the Courtreferredwas neithertestimonialnor documentaryin
characterbut came from a "Brandeis"-typebrief filed directly in the Supreme Court by the
National Association for the Advancementof Colored People.
Id. at 678 (citationsomitted).Althoughthe judge's characterizationof the factualand evidentiarybasis of
Brownseems fair, his analysis needs some such distinctionas thatbetween "legislative"and "adjudicative"
facts-a distinction that did not become common until 1975, when Rule 201 of the Federal Rules of
Evidence made it salient. See FED.R. EvID. 201(a) advisory committee'snote. Such a distinctionwould
have made his argumentfar less plausible,for it was the failureto distinguishthese types of facts thatgave
this stare decisis argument whatever superficial plausibility it had. Brown's holding rested on a finding not
of adjudicativefacts, but ratherof legislative facts. While it is true that adjudicativefacts cannot bind
nonparties, legislative facts are much more likely, as a matter of institutional practice, to do so, as Brown
itself shows. (Whetherand to whatextent a courtshouldbind otherpartiesto its findingsof legislative facts
is a separatequestion, a question on which Brownprovides no guidance.)
111. See Stell, 220 F. Supp. at 676-77.
112. Id. at 678.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1562 The Yale Law Journal [Vol. 107: 1535

present case as having somewhat stronger indicia of truth than that on which
the findings of potential injury were made in Brown."113
On appeal, the Fifth Circuit slapped down the lower court's procedural and
evidentiary insurrection."' One can well understand the likely urgency with
which that court, as the faithful agent of the Supreme Court, must have felt the
need to quell this kind of rebellion before it spread. But is there a principled
basis for the Brown Court to have created, as it apparently intended to create
and as courts like the Fifth Circuit clearly understood it to intend, what the
Stell district court aptly termed "a conclusive presumption of injury to Negro
students by reason of segregation"?"5 If the Court really wanted the
authority of science, was it not obliged, as a matter of principle, to allow
science to reach modest conclusions, tentatively held and ever open to revision,
that are among the most notable characteristics of empirical science-
characteristics probably even more important for social science than for natural
empirical science? By resting so heavily on empirical science, nay, on the least
stable kind of empirical science, the Court opened itself up to the charge of
fiat, of power, rather than reason, in shutting down efforts by the polity to
revisit the scientifically informed fact at issue."6 The very modesty of truly
scientific conclusions does not warrant so strong a result as the Court dictated
with its irrebuttable "inherently." Truly scientific results would seem not only
to permit, but also to invite, if not to require, fresh reexamination.
Many scholars have been distressed at the Court's use of factual evidence
in Brown. Some contend that the Court's decision to "constitutionalize" a
putatively scientific result is deeply problematic because "attaching
constitutional meaning to scientific opinion, even when scientists are in
consensus, condemns the Constitution to fluctuations in meaning as scientific
knowledge changes."117 That admonition is not wholly without merit, but it

113. Id. at 680.

114. The Fifth Circuitnoted in its opinion:
We do not readthe majorpremiseof the decision of the SupremeCourtin the firstBrowncase
as being limited to the facts of the cases therepresented.We read it as proscribingsegregation
in the public educationprocesson the statedgroundthatseparatebut equal schools for the races
were inherentlyunequal.This being our interpretation of the teachingof thatdecision, it follows
that it would be entirely inappropriatefor it to be rejectedor obviatedby this court.
Stell v. Savannah-Chatham CountyBd. of Educ., 333 F.2d 55, 61 (5th Cir. 1964), cert. denied, 379 U.S.
933 (1964). After the SupremeCourtdenied the state's petitionfor certiorari,the districtcourt on remand
granted an injunction against the school board's operation of a segregated school system and further
enjoined the board:
[C]onsistentlywith and in accordancewith Brownand the decision of the Courtof Appeals in
this case school children may be assigned to particularschools "on a basis of intelligence,
achievement,or other aptitudesupon a uniformlyadministeredprogram"providedrace is not
a factor in making the assignment.
Stell v. Savannah-Chatham CountyBd. of Educ., 255 F. Supp. 88, 94 (S.D. Ga. 1966) (citationomitted).
115. Stell, 220 F. Supp. at 668 (emphasisadded).
116. For a classic complaintabout courts' use of fiat ratherthan reason in attemptingto justify their
discussions, see Lon L. Fuller,Reason and Fiat in Case Law, 59 HARV. L. REv. 376 (1946).
117. David L. Faigman, "Normative Constitutional Fact-Finding": Exploring the Empirical
Componentof ConstitutionalInterpretation,139 U. PA. L. REv. 541, 573 (1991) (footnote omitted).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1563

is not easy to discern exactly what its merit is. Ronald Dworkin is also among
those who have worried about the relation between empirical judgment and
normative constitutional judgment, and his version of the concern is instructive.
His basic solution to the problem-in an argument that presages his later and
more expressly "hermeneutic" work-was to recharacterize the nature of the
evidence in Brown so that the necessary evidentiary judgment was not one of
science, but was rather an "interpretive" judgment that is presumably more
comfortably within the scope of judicial prowess. As Dworkin put it:

"We don't need evidence for the proposition that segregation is an

insult to the Black community-we know it; we know it the way we
know that a cold causes snuffles." It is not that we don't need to
know it nor that there isn't something there to know. There is a fact
of the matter, namely that segregation is an insult, but we need no
evidence for that fact-we just know it. It's an interpretive fact."18

Dworkin's analogy is inapt. Surely our knowledge of the relation between

a cold and "snuffles" rests on causal judgments that are empirical. Moreover,
the kind of judgment the Court made in Brown does seem inevitably to be an
empirical one, for which evidence is needed, and regarding which expertise is,
at least in principle, available."9 The difficulty of making such judgments
reliably cannot be wished away by recharacterizing them as "2interpretive.20

118. RonaldDworkin,Social Sciences and ConstitutionalRights-The Consequencesof Uncertainty,

6 J.L. & EDUC.3, 5 (1977) (paraphrasingEdmondCahn, Jurisprudence,30 N.Y.U. L. REv. 150, 157-58
(1955)).
119. Dworkin focuses on a "finding"slightly differentfrom that at issue in Brown.Brown's finding
was that segregationcaused psychological harm.Dworkin'sissue is whetherit was an "insult."Not every
insult causes psychological harm,and when an insult does so is a questionof empiricalpsychological fact.
Nevertheless, his question whethera given social gesture is an insult in a particularcommunityor to a
given person is also thoroughlyempirical.
120. Dworkin'sanalysisis an early versionof argumentsaboutso-called "socialmeaning"(whatother
kind is there?) that are becoming popular among legal academics. See, e.g., Lawrence Lessig, The
Regulation of Social Meaning, 62 U. CHI. L. REv. 943 (1995). Like Dworkin's approachto general
jurisprudentialissues, and like much recentwork in literaryand hermeneutictheory,this approachis much
absorbedby the effort to linguistify and textualizefeaturesof legal and social experiencethat are readily
explained in nonlinguisticterms. According to one of its recent expositors, Professor Lessig, the social
meaning approachfocuses on "the semiotic content attachedto variousactions, or inactions,or statuses,
within a particularcontext,"id. at 951, for the purposeof assessing whether(1) collective action problems
exist that make it difficult to "reconstruct"social meaningsthat have been "constructed";and (2) whether
and how such collective action problemsas do exist should be addressedby privateor public agents, see
id. On this approach,for example, when an action creates a stigma, the stigma is a social meaning, and
when a gesture is an insult, the insult is a social meaning. See id. Moreover, a stigmatizing socially
meaningful act like discriminatingon the basis of race in a publicly visible way can create a collective
action problem for stigmatizerswho would rathernot behave in such a way, other things equal, but who
feel compelled to do so by a social meaning that they cannot individuallycontrol. Lessig claims that this
predicamentafflicted many employers in the South prior to passage of the Civil Rights Act, which then,
Leviathan-like,solved this collective action problem.See id. at 965-67.
An enduring question about all such approachesis to what extent theories of linguistic meaning
narrowlyconceived can usefully be extended to explain phenomenathat seem linguistic only in a loose
or-dare I say it-metaphorical sense. An answerto that question,however, awaits discussion elsewhere.
For now, observe that such a theory easily lends itself (indeed, these very examples lend themselves) to

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1564 The Yale Law Journal [Vol. 107: 1535

But even if Dworkin's and similar efforts cannot convert empirical questions
into (allegedly) nonempirical "interpretive" or "semiotic" ones, the concern
with "constitutionalizing" putatively scientific empirical results is a real one.
Let us come back directly to the closely related problems that are my chief
concerns in this Article, those of institutional competence and justified
epistemic deference-whether courts, using rules of evidence, procedure, and
other institutional and doctrinal devices, are competent to effect a transfer of
experts' justified beliefs to judges or juries in a way that is sufficiently
legitimate for legal decisionmaking.
Two questions present themselves about the Court's use of expert
information in Brown. First, did the Court really face no serious problem of
competing experts of the sort that threatens the legitimacy of nonexpert judge
and jury decisions under the Daubert regime? Second, is a nonexpert court
capable of discerning the degree of instability in a necessarily tentative
empirical finding-that is, the degree of epistemic repose that the finding
warrants12-and of discerning it sufficiently for the purpose of bringing
about the amount of legal repose that relevant scientific factual findings
warrant?'22
As to the first question, it may be that Brown lucked into a factual finding
that really was supported by the evidence and by the opinion of a majority of
reputable experts (presumably these are not wholly independent variables!),
despite the contrary evidence vigorously adduced by the lawyers in Stell. Such
a degree of support among those we take to be experts triggers in us what I
shall refer to as "the Frye response": If a critical mass of duly recognized
experts support a claim, that claim is very, very likely to be true. (Whether that
response is also epistemically warranted is the issue I take up in more detail
in a later discussion.123 It is crucial to keep in mind that truth alone is no

a Dworkin-like analysis of Brown. Perhaps what Brown was really addressing was whether there was a
"social meaning" of state-mandated segregation, and whether, if so, it created a psychological harm to black
schoolchildren. Dworkin's suggested answer to the first question is, yes, the social meaning was that of
"insult." But even through the "social meaning" lens, there are still fundamentally empirical questions, pace
Dworkin, to be asked about whether in fact a given act or gesture is insulting and stigmatizing, and even
if so, whether that insult or stigma in fact causes psychological harm. Thus, the social meaning
interpretation of Brown cannot obviate, and instead merely restates or relocates, the problem of how courts
can reliably acquire empirical information of this sort.
121. Daubert gives a nice example of what I am referring to here as epistemic repose:
[W]e do not read the requirements of Rule 702 to apply specially or exclusively to
unconventional evidence. Of course, well-established propositions are less likely to be
challenged than those that are novel, and they are more handily defended. Indeed, theories that
are so firmly established as to have attained the status of scientific law, such as the laws of
thermodynamics, properly are subject to judicial notice under Federal Rules of Evidence 201.
Daubert v. Merrell Dow Pharms. Inc., 509 U.S. 579, 593 n.l 1 (1993).
122. The "legal repose" at issue here is both internal to a particular case and external to it. Doctrines
like res judicata and the pragmatic relevance requirement of Federal Rule of Evidence 403 guide judges
in effecting repose within a case. See FED. R. EVID. 403. Judges' use of doctrines of stare decisis help
effect legal repose across cases.
123. See infra Section V.C.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1565

guarantor of epistemic warrant.) But even if the Brown Court did luck into
such a finding, that is precisely the point: The Court lucked into it. It was not
for any skill on the part of the Court in sorting through dozens of studies,
many of which surely relied on statistical modeling and other complex
methodologies, that the Court was able to discern the truth of the psychological
claim. Nor is a court in general competent to do so. Thus, in many other cases
in which social-scientific evidence is relevant, perhaps nearly dispositive,
courts will be in the same position they are in with regard to results in the
hard sciences: placed at the mercy of competing experts. Indeed, other cases
in which psychological evidence is powerfully relevant, and which expose
courts' limited epistemic competence come readily to mind; consider the
increasing reliance by litigants on putatively expert testimony regarding various
syndromes, including post-traumatic stress disorder and battered woman or
battered child syndrome.'24
Consider now the second question, the question of how much legal repose
is warranted by the degree of epistemic repose in an empirical study. Courts
do not appear to be generally competent to make the subtle judgments required
to balance appropriately epistemic and legal repose. To make such judgments,
a court should be capable of assessing rationally the extent to which empirical
data support a conclusion, which would in turn require the judge (or judge plus
jury) to understand formal models, experimental design, and so on. The
epistemic problem presented by the questions of the proper degree of epistemic
and legal repose might even be framed as a problem of "diachronically"
competing experts.125 Even if most experts agree now, at what point in the
future should courts be open to challenges to what is now agreed? In a way,
that is the question that Stell posed to Brown, and neither the Fifth Circuit in
Stell nor the Supreme Court in Brown itself provided any guidance.
The real problem with cases like Brown, no less than in cases like (and
cases decided pursuant to) Daubert, is that scientifically nonexpert judges and
juries do not seem institutionally competent to assess expert scientific evidence
rationally. In a Daubert-type case, the main problem will be the battle of
experts that is waged before spectators who are for the most part not
competent even to understand, much less to apply in a nonarbitrary manner,
that intellectual contest's rules. In Brown, there are likely to be two significant
problems: the capacity of judges and juries to assess competing scientific
testimony, and the capacity of judges to decide, in a principled way, how to
balance the inherent instability of empirical results against the law's desire for
finality, repose, stability, and incremental and conservative evolution. As I
argue below in greater detail (but surely the claim, as vague as it is here, has
intuitive force), legal decisionmakers should make decisions that are not

124. See MUELLER & LAIRD, supra note 19, at 729-41.

125. For extended discussion of "competition" in experts' testimony, see infra Section IV.C.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1566 The Yale Law Journal [Vol. 107: 1535

fundamentally arbitrary from an epistemic point of view.'26 Failure to do so

threatens the legitimacy of the decision from a legal point of view. The
reasoning in both Brown and Daubert inspires little confidence that courts are
capable of overcoming this threat.

II. AxIOLOGY AND "POINT OF VIEW" IN THE ANALYSIS OF

EPISTEMICDEFERENCE

A. Central Relation: Epistemic Deference by "Practical" (Legal) Reasoners

to "Theoretical" (Scientific) Experts

Brown and Daubert are tips of the iceberg. Questions to which scientific
theories and methods are reasonably likely to have answers that are both
material and relevant (in the older common law sense)127 are raised in an
increasingly large percentage of cases at civil and criminal law. Among the
scientific theories and methods that have been prominent in litigation in recent
years, for example, are the use of genetic theories and laboratory methods to
assess "DNA fingerprint" evidence; findings by psychologists regarding the
veracity of eyewitness testimony; and complex medical and epidemiological
studies regarding the harm that food, drugs, and technologies have caused or
can cause.
Legal and political systems around the world cede legal decisionmaking
authority regarding such questions to scientifically nonexpert decisionmakers
(those who are largely untrained in any field of scientific research) to varying
degrees. Some systems, like that of the United States, give nonexperts the
ultimate say on a great many legal issues to which scientific information is
reasonably likely to be relevant and material. Several Continental systems are
far less inclined to give these science-imbued decisions to nonexperts.128

126. See infra Part VII.

127. See supra note 19 and accompanyingtext.
128. France is often mentioned as the chief exemplar of a political and legal system that gives
decisions to "technocrats"who have receivedextensivetrainingin (interalia) science, but who also exercise
a great deal of bureaucraticgovernmentalpower withoutany real oversight by democraticinstitutions.It
is probablynot quite rightto conclude that this is a system in which scientists make many of the decisions
thatjudges andjuries would make in the United States.It appearsthat,althoughthe elite Frenchtechnocrat
does typically get trainingin mathematics,physics, and chemistry (as well as economics, law, decision
theory,and administration),he is specifically trainedto be a generalist manager, whose only specialty is
the multipurposeskill of governing, administering,and making organizationaldecisions-a skill that is
thoughtto requiresufficient competencewith science to comprehendits use in policy decisions. Thus, for
example, students of one of the elite trainingacademies have been told routinely that "'[tihe scientific
trainingyou receive will not give you the knowledgein any branchthatthe specialistshave, but it will give
you the aptitudesand the methodssuch as to allow you to be on top of everything."'EZRAN. SULEIMAN,
ELITESIN FRENCHSOCIETY:THE POLITICSOF SURVIVAL166 (1978) (quoting language addressedto an
entering class of the Ecole Polytechnique,in FREDERICBON & MICHELA. BURNIER,LES NOUVEAUX
INTELLECTUELS 112 (1971)). For discussionof the issue, see JOHNARDAGH,FRANCEIN THE 1980s, at 82-
92 (1982); and SULEIMAN,supra, at 158-92. Even if these technocrats are not scientists in their own right,
it does seem clear that they are far better trained in scientific matters than the average American juror or
judge. It also seems clear that government officials trained in this way make many far-reaching decisions

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1567

This Article constructs a model of the reasoning process that nonexpert

judges, juries, and other legal reasoners use in the effort to reach justified legal
decisions when those legal reasoners rely on-defer to-expert scientific
evidence. In developing this model, I hope to provide a philosophical
explanation of how this particular reasoning process works, which in turn can
advance our thinking about how it might be justified and whether it is justified,
as currently practiced in American legal institutions. Although my chief
explanatory target is the process of deference by legal reasoners to experts,
such deference is only one species of a broader generic phenomenon: the
deference by practical reasoners (those who reason about what they ought to
do) to expert theoretical reasoners (those whose expertise lies in some area of
reasoning about how the world is). I shall refer to this as "practical epistemic
deference." In providing the model of practical epistemic deference, I use
examples from both legal and nonlegal practical reasoning. The legal examples
will loom largest, however, in part because the express institutionalized rules
that guide the legal system in this kind of deference-chiefly, rules of evidence
and procedure-make this reasoning process in law more amenable to analysis
than the comparable but less explicitly regulated process of nonlegal deferential
practical reasoning.
Another qualification is equally important. Although I spend some time
discussing and relying on a distinction between practical and theoretical
reasoning, I limit my analysis in this Article to scientific theoretical reasoning,
albeit with an expansive conception of the scientific, embracing "hard" and
"soft" empirical science as well as the demonstrative sciences of mathematics
and logic-all of which can be and have been the object of epistemic
deference by legal reasoners to experts. I limit my analysis in this way for two
main reasons. First, courts and scholars concerned with the problem of
epistemic deference have themselves focused on scientific knowledge as a
distinct category of expert evidence. For example, Daubert expressly limited
its analysis to the admissibility of scientific expert testimony.'29 Left out of
the Court's analysis was any discussion of the admissibility of nonscientific
expert testimony, such as the testimony a police officer might give about the
motives and habits of drug dealers. The second reason is heuristic. My
intuitions about the problems associated with nonexperts' deferring to experts
are clearest with regard to scientific evidence. If we can make a significant
advance in understanding the structure of justification that frames nonexperts'
reasoning about scientific expert testimony, we may use that as a heuristic
lever to gain clarity about deference by nonexperts to other types of experts-a
task I leave for later work.

involving science and technology that are given to judges and juries in the United States, particularly in
torts cases.
129. See Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579 (1993).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1568 The Yale Law Journal [Vol. 107: 1535

For the purposes of my inquiry here, assume that there is a cogent and
important distinction between practical and theoretical reasoning. I rely on that
assumption to explore and explicate epistemic deference by legal reasoners to
scientific experts. I characterize these as distinct "points of view" and begin
by explaining what a "point of view" is and how I use the concept in my
analysis.

B. Reasoning from a "Point of View"

Many philosophers have invoked the concept of the "point of view."

Oliver Wendell Holmes's "bad man" is an analytical device that offered a
perspective, a point of view, on the law (the point of view, perhaps among
others, of the lawyer).'30 H.L.A. Hart extended this basic idea and made it
a central feature of his positivist account of law: the distinction of the
"internal" and "external" attitudes-points of view-toward legal rules.'3'
Other positivists, such as Joseph Raz and Hans Kelsen, have relied heavily on
the idea of a point of view in explaining the nature of law.'32 Dworkin, too,
has relied repeatedly on the concept of the point of view, distinguishing the
external point of view (that "of the sociologist or the historian") from the
internal point of view of the participant in a social practice like
adjudication.'33 Dworkin declares that his own analysis of law is from the
"internal," "participants"' viewpoint, from which he "tries to grasp the
argumentative character of our legal practice by joining that practice and
struggling with the issues of soundness and truth participants face."'34
Dworkin has also argued that legal interpreters seek to put the legal materials
they interpret in their "best light," best "from the standpoint of political
morality."'35From a nonpositivist perspective quite distinct from Dworkin's,
Ernest Weinrib has articulated a "formalist" legal theory that insists on the
vital importance of taking an "immanent" point of view.'36 According to

130. See OLIVER WENDELL HOLMES, ThePath of the Law, in COLLECTED LEGAL PAPERS 167,passim
(1920); see also William Twining, The Bad Man Revisited,58 CORNELLL. REV. 275 (1973).
131. See H.L.A. HART, THE CONCEPT OFLAW 89-91, 242-43, 254 (2d ed. 1994).
132. See JOSEPHRAZ, THE AUTHORITYOF LAW: ESSAYS ON LAW AND MORALITY137-45 (1979)
[hereinafterRAz, THE AUTHORITYOF LAW] (discussing Kelsen); JOSEPHRAZ, PRAcrICAL REASONAND
NORMS 170-77 (1975) [hereinafterRAZ, PRAcTICALREASON].
133. DWORKIN,supra note 3, at 14.
134. Id.
135. Id. at 255-56. Dworkin maintains:
Hardcases arise, for any judge, when his thresholdtest does not discriminatebetween two or
more interpretationsof some statute or line of cases. Then he must choose between eligible
interpretations by asking which shows the community's structure of institutions and
decisions-its public standardsas a whole-in a better light from the standpointof political
morality.His own moral and political convictions are now directly engaged.
Id. (emphasis added).
136. See ErnestJ. Weinrib,TheJurisprudenceof Legal Formalism,16 HARV. J.L. & PUB.POL'Y583,
passim (1993) [hereinafterWeinrib,Jurisprudenceof Legal Formalism];see also ErnestJ. Weinrib,Legal
Formalism: On the ImmanentRationalityof Law, 97 YALE L.J. 949 (1988).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1569

Weinrib, "formalism represents the law's aspiration to be an immanently

intelligible normative practice. Immanence bespeaks a standpoint that is
internal to law" and thus has "a critical standpoint, but one that emerges from
the law's own aspirations."'37 One hotly contested claim in a branch of
recent legal academic writing is that there is a distinct "women's point of
view" or "black point of view."'38
The concept of the point of view has played an important role for many
nonlegal philosophers, too. W.V. Quine, for example, has distinguished the
"epistemic" and the "ontological" points of view to explain how his complexly
related theses of underdetermination and indeterminacy do not commit him to
self-refuting relativism.'39 John Rawls has suggested the importance of the
notion of "point of view" in distinguishing the roles and goals of the legislator
and the judge.'40 Thomas Nagel has devoted an entire book to

a single problem: how to combine the perspective of a particular

person inside the world with an objective view of that same world,
the person and his viewpoint included. It is a problem that faces every
creature with the impulse and the capacity to transcend its particular
point of view and to conceive of the world as a whole.14'

Laurence Bonjour has argued the importance of identifying the epistemic point
of view, maintaining that "[t]he distinguishing characteristic of epistemic

137. Weinrib,Jurisprudenceof Legal Formalism,supra note 136, at 591-93 (emphasisadded).

138. There is good reasonto doubtwhetherit is epistemicallyor morallyvaluableto believe thatthere
are such points of view; that is, whetherfrom eitheran epistemologicalor a moralpoint of view, we ought
to think that there is such a thing as a woman's or a black person'spoint of view. For critical discussion
of this issue, see Scott Brewer,Pragmatism,Oppression,and the Flight to Substance,63 S. CAL. L. REv.
1753 (1990).
139. Quine argues:
From among the various conceptual schemes best suited to these various pursuits,
one-the phenomenalistic-claims epistemological priority. Viewed from within the
phenomenalisticconceptualscheme, the ontologiesof physicalobjectsand mathematicalobjects
are myths. The qualityof myth,however,is relative;relative,in this case, to the epistemological
point of view. This point of view is one among various, correspondingto one among our
various interestsand purposes.
QUINE, supra note 63, at 19 (emphasisadded). Quine continues:
For my partI do, qua lay physicist, believe in physical objects and not in Homer's gods;
and I consider it a scientific errorto believe otherwise.But in point of epistemologicalfooting
the physical objects and the gods differ only in degree and not in kind.... The myth of
physical objects is epistemologicallysuperiorto most in thatit has provedmoreefficacious than
other myths as a device for workinga manageablestructureinto the flux of experience.
Id. at 44 (emphasis added). The distinctionof the epistemologicalfrom the ontological points of view is
centralto Quine's projectof "naturalizedepistemology."See, e.g., W.V. QUINE, WORD AND OBJEcr 22-23
(1960). As Quine explains it, when we speak of underdetermination of a physical theory,we are speaking
"fromthe standpointof a descriptionof the theory-buildingprocess,"and noting that in buildingtheories,
"physicalhumansubjects"go (and inevitablygo) beyond the sensory evidence on which the theories are
built;most generally,the projectin which scientistsareengaged is the ontologicalprojectthe goal of which
is "painstakinglyto surmisewhat realityis like." Id. at 22; see also id. ("Everythingto which we concede
existence . . . is . . . real from the standpoint of the theory being built." (emphasis added)).
140. See John Rawls, Two Conceptsof Rules, 64 PHIL. REv. 3, 6-7 (1955).
141. THOMAS NAGEL, THE VIEW FROM NOWHERE 3 (1986).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1570 The Yale Law Journal [Vol. 107: 1535

justification is . .. its essential or internal relation to the cognitive goal of

truth."'42
Properly understood, the concept of a "point of view" will prove quite
useful for the analysis of practical epistemic deference-deference by practical
reasoners to scientific experts. Thus I now explicate this concept in a way that
captures what is common to virtually all the aforementioned philosophical uses
of the idea and indicates how that concept may advance the analysis of
practical epistemic deference. Some observations about (more or less) common
usage will usefully frame the issue. One might be said literally to have a point
of view, that is, actually to occupy some position in space that gives one a
particular visual vantage. On the forest floor, one might see only trees; from
a point atop a mountain, one might see the forest, and not only the trees. From
a bird's-eye view (say, from an airplane), one might see the shape of a lake;
and from an astronaut's-eye view, the shape of the earth. It is also common to
use the term "point of view" to speak of expertise. An expert witness might
tell a jury or judge what the facts are from the point of view of a biologist or
a chemist or a ballistician or a psychiatrist-indeed, the point of view of the
expert scientist will be of central concern for my later analysis of practical
epistemic deference. The facts that are salient from these expert points of view
will by no means always be the same as the facts that are salient from some
practical point of view, such as a legal or moral point of view.
One may also refer to the point of view of a particular type of actor in an
institutional or other social setting-the point of view of a legislator or a
judge, a lawyer or a citizen, a president or a "bad man," a parent or a child,
a professor or a student. One may also refer to the point of view not of a given
type of institutional or other social actor, but of the enterprise in which the
actor acts-the enterprise that gives a judge, parent, or teacher his identity as
that kind of actor. This might be understood as the point of view of an
enterprise, an enterprise in which particularmethods of analysis are chosen and
used to serve specified cognitive goals. Examples of such enterprises include
systems of moral reasoning (on a cognitivist account of morality, at least, this
yields the "moral point of view"), philosophical reasoning (the philosophical
point of view), systems of reasoning in support of business objectives (the
business point of view), and many others that are familiar (the military point
of view, the economic point of view, the religious point of view, and so on).
The "enterprise" conception of point of view is, I suggest, the common
thread running through all the notions of point of view mentioned above, both
the "common" and the more reflectively philosophical. The superficially
varying conceptions of point of view noted above are reducible to the

142. LAURENCEBONJOUR,THE STRUCTUREOF EMPIRICALKNOWLEDGE8 (1985). The passage

continues: "It follows that one's cognitive endeavorsare epistemicallyjustified only if and to the extent
thatthey are aimed at this goal, which means very roughlythatone accepts all and only those beliefs which
one has good reason to think are true."Id.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1571

enterprise conception by virtue of the close connection each use of point of

view has to claims of justification, in that claims of justification are advanced
and measured by reference to the cognitive goals of an enterprise and the
methods the enterprise uses to serve those goals. To explain this point slightly
differently, in each use of point of view noted above, the concept of point of
view is invoked to justify some claim-either a claim about what we ought to
believe (a theoretical claim) or how we ought to act (a practical claim).
Each claim is thus implicitly a claim that a certain belief is justified, and
that its justification comes in significant part from the method the epistemic
"enterprise" uses to produce it. If you are justified in claiming that a lake is
rectangular in shape, it is because you have a bird's eye (or some other
reliable) point of view on it (or are reliably relying on someone who does). A
geneticist might be justified in believing, on the basis of "DNA fingerprint"
evidence, that a defendant was present at the scene of a crime, because the
geneticist has the appropriate expert point of view. Because she is a lawyer (or
other legally trained person), and so has the lawyer's point of view, a person
might be justified in her belief that her friend ought not sign some contract
(given, of course, that she has sufficient information about the friend's interests
and desires). And a military commander's claim that a particular bombing
mission was the right thing to have done might be correct from a military point
of view although highly questionable from either a legal or a moral point of
view. Of course, simply identifying the general point of view of an enterprise
does not by itself answer the following question: What are the specific aims
of the enterprise for theorists who recognize themselves as pursuing the same
generic enterprise, but who often disagree about what are the proper specific
aims of the enterprise? Such disagreements are a principal source (but not the
only source) of the difference among theories within an enterprise.
Further refinement of the "enterprise" conception of a point of view (the
conception to which others are reducible) is useful. What is distinctive of an
enterprise on whose behalf a participant in that enterprise makes a justificatory
claim is that the claim was the product of a distinct method, and that the
method, in turn, is chosen by enterprise participants to achieve distinct
cognitive aims. The military point of view differs from the moral point of
view, for example, because the enterprise of assessing a given action from each
point of view differs according to the distinct methods and cognitive aims of
military and moral analysis.
We may refer to this conception of point of view that focuses on the
distinctive intellectual aims and methods of a given enterprise as an
"axiological" conception. The axiological conception will help us map out the
relations between the intellectual framework in which a judge or jury reasons
about a given legal dispute (a legal point of view) and the rather distinct
framework within which an expert scientist reasons about facts that are
material and relevant to legal decisions. In offering this "enterprise" and

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1572 The Yale Law Journal [Vol. 107: 1535

"axiological" conception of the point of view, I borrow from Larry Laudan's

clarifying analysis of scientific reasoning.'43 Laudan presents a model of
scientific reasoning that identifies three distinct levels of analysis: the
"factual," the "methodological," and the "axiological." At the factual level are
"all manner of claims about what there is in the world, including theoretical
and unobservable entities."'44At the methodological level are methodological
rules that scientists share.'45There is great variety in the scope and content
of these rules, ranging from the rather vague ("avoid ad hoc explanations,"
''accept only independently testable results," "formulate testable and simple
hypotheses," "prefer the results of double-blind to single-blind experiments")
to the very precise ("make sure to calibrate instrument x against standard
y",).146 It is the application of methods to experimental data (pursuant to
methodological rules) that warrants a scientist's factual assertions; by applying
methodological rules in this way, scientists who accept those rules are often
able to resolve factual disagreements.
Laudan's model of scientific reasoning identifies a third crucial level of
analysis, the "axiological" level.'47 At this level are statements, often in the
form of rules, of the "cognitive aims" of science. For example, a scientist
ought to use methods for generating factual beliefs that achieve empirical
accuracy; or achieve coherence with the set of currently held theories and
beliefs; or produce the simplest theory among those theories that explain the
data equally well; or produce a predictively fertile theory; or produce a theory
that possesses a high degree of generality and breadth of scope.'48 Although

143. See LARRY LAUDAN, SCIENCEAND VALUES: THE AIMS OF SCIENCEAND THEIR ROLE IN
SCIENTIFICDEBATE(1984).
144. Id. at 23.
145. See id.
146. Id.
147. See id.
148. See id. at 26. Thomas Kuhn identifies the following aims of scientific inquiry: accuracy
(deducibleconsequencesshould be in agreementwith experiments);consistency (both internallyand with
other currentlyaccepted theories); simplicity (the explanationshould presupposethe existence of fewer
ratherthan more entities);fruitfulness(disclose new phenomenaor previouslyunnotedrelationships);and
broadscope (consequencesshould extend beyond initial observations,laws, or subtheories).See THOMAS
S. KUHN, Objectivity, Value Judgment, and Theory Choice, in THE ESSENTIAL TENSION: SELECTED STUDIES
IN SCIENTIFICTRADITIONAND CHANGE320, 321-22 (1977). Quine and J.S. Ullian identify and describe
"five virtueswhich count towardplausibility,and which a hypothesismay enjoy in varyingdegrees."W.V.
QUINE & J.S. ULLIAN, THE WEB OF BELIEF43 (1970). These virtues are conservatism (the hypothesis
shouldconservepriorbeliefs because"[tgheplausibilityof a hypothesisvariesinverselywith the plausibility
of the prior beliefs that it disallows," id. at 44); generality ("[tihe plausibility of a hypothesis depends
largely on how compatiblethe hypothesisis with our being observersplaced at randomin the world,"id.);
simplicity (that,when choosing among hypothesesequalin all otherrespects,the moreplausiblehypothesis
is the simpler one, see id. at 45); refutability(a hypothesis neitherconfirms nor predicts anythingunless
"[slome imaginableevent, recognizableif it occurs . . . suffice[s] to refute it," id. at 50); and modesty (a
stronger "conservatism"demandingthat the less "extravagant"hypothesis be preferredto those that are
more so, other things being equal, id. at 51).
One might treat as axiological values (aims) the values that Kuhn,on the one hand, and Quine and
Ullian, on the other, identify, but one might also treat some of them as methodologicalrules (methods).
Laudanhimself does not drawa sharpline betweenmethodsand aims. See, e.g., LAUDAN, supra note 143,
at 31. One quite plausible explanationfor this is that no sharp line can be drawn between the concepts

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1998] Scientific Expert Testimony 1573

such cognitive norms are often vague,149 they are not so vague as to be
without action-guiding or belief-guiding content.
Suffusing his discussion with examples from the history of science,
Laudan uses his tripartite model of scientific reasoning to argue against both
"arch-rationalist" and "arch-undeterminationist"philosophers of science. The
former are those who believe that there is "an algorithm or set of algorithms
which would permit any impartial observer to judge the degree to which a
certain body of data rendered different explanations of those data true or false,
probable or improbable."'50 The latter are those who believe that because
"we have no perfectly general logic of confirmation or comprehensive theory
of evidence," and because many scientific rules (methodological and
axiological) are vague or ambiguous or highly general, "the application of
shared scientific [methodological] rules or [axiological] values to a specific
choice situation will always be (or always has been) unavailing."'5' (Thomas
Kuhn is Laudan's chief target.)
Against the rationalists, Laudan advances several arguments and
observations (well known in the work of Norwood Hanson, Paul Feyerabend,
Quine, Kuhn, and others). For example, Laudan observes that methodological
rules, though shared by scientists, may still underdetermine preferences among
competing factual beliefs, and shared axiological goals may similarly
underdetermine a preference among competing methodological rules.
Furthermore, even among scientists who accept the same set of axiological
aims (e.g., coherence, simplicity, empirical accuracy), that set will
underdetermine a preference among competing methodological rules whenever
competing methodological rules promote or thwart different members of the
set of axiological goals in different ways. To choose among competing
methodological rules that satisfy some axiological goals but thwart others,
scientists must weigh those goals, and since they might weigh them differently,
even a shared bundle of goals may underdetermine the choice of
methodological rules. While he criticizes the rationalists, Laudan also deploys

"method"and "aim,"and that (as with all vague concepts) the line between them is to be drawnon the
basis of holistic considerations. Consistent with this explanation is the possibility that there is one
fundamentalaxiological aim that all other aims and methodssubserve.For science, a likely candidatefor
such an overall and overridingand supersedingaxiological norm is a norm of truthor accuracy,such as,
"Choose those aims, methods,and beliefs that are most likely to producetrue or accurateaccountsof the
world." Cf. BONJOUR,supra note 142, at 7 ("Whatmakes us cognitive beings at all is our capacity for
beliefs, and the goal of our distinctivelycognitive endeavorsis truth:we want our beliefs to correctlyand
accuratelydepict the world."). Note that if there were such a norm, it would play a structuralrole for
judgments of scientific validity analogous to the role that, on Kelsen's account, the Grundnormplays for
judgments of legal validity. See RAZ,THE AUTHORITYOF LAW,supra note 132, at 122-45 (discussing
Kelsen's "basic norm").
149. Every empirical concept is "open-textured,"i.e., has the possibility of being vague in some
particularcircumstance.See Brewer,supra note 15, at 993-96. If, with Quine, we abandonone of the two
dogmas of empiricism,we would acknowledgethat all concepts, even mathematicaland logical concepts,
are open-textured.See QUINE, supra note 63, at 43.
150. LAUDAN, supra note 143, at 5.
151. Id. at 32.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1574 The Yale Law Journal [Vol. 107: 1535

several arguments against the underdeterminationists,152 such as that while

a methodological rule may underdetermine choice among competing factual-
level beliefs in the abstract, "the same rule may still unambiguously dictate a
comparative preference among extant alternatives"153-that is, it can
determine preference among competing beliefs that are under active
consideration even while it underdetermines beliefs overall.'54
I maintain that this basic axiological structure is generalizable from
scientific theory to any intellectual discipline that has a distinct intellectual
method. I need not, and do not, endorse the conclusion that Laudan has
achieved a complete analysis of scientific rationality. Surely Laudan's analysis
can be questioned for its basic assumption that instrumental rationality is the
essence of scientific rationality-an assumption clearly reflected in Laudan's
dissection of science into an enterprise driven by the selection of aims and
methods for achieving them.155 But even if other conceptions of rationality
fill out the picture, it does seem that a structure of the sort that Laudan
identifies comprises a very significant part of the story of scientific rationality.
I also believe that, suitably generalized, this structure provides both a full
explication of the concept of a point of view and a deep insight into the nature
of any intellectual discipline that uses a distinctive method to justify claims.
Another virtue of Laudan's model is its ready amenability to the pattern
of "reflective adjustment" that I have emphasized elsewhere.156 The concept
of reflective adjustment provides a powerful explanation of the structure of
justificatory reasoning in a wide variety of fields (including not only analogical
reasoning, but also moral theory, logic, and epistemology). Laudan's
axiological conception extends this model of justification to the realm of
scientific reasoning. He argues, contrary to claims by Rudolf Carnap, Hans
Reichenbach, Thomas Kuhn, and others, that scientific justification cannot be
adequately explained as a linear hierarchy in which a scientist justifies fact
claims by reference to methods and justifies choice of method by reference to
aims.157 Although on Laudan's model, facts, methods, and aims all play a
critical role, they cannot be arrayed on a simple linear hierarchy in which
factual claims are constrained only by methodological rules, and

152. Loud analogiesto consensusand dissensusformationin both legal theoryand legal interpretation
ring in one's ears. Legal theoryhas certainlyhad its arch-rationalists(legal process theorists,perhaps,and
Dworkinians), its arch-underdeterminationists (some Realists, some Crits), and its Laudan-likecentrists
(such as H.L.A. Hart).
153. Id. at 29.
154. For example, in biology, the rules and evidence do not "establishthe unique correctnessof
evolutionarytheory,"but they do rule out several creationistclaims from the permissiblerealm "andthus
provide a warrantfor a rational preference for evolutionary over creationist biology." Id. Of course,
Laudan'sclaim here can be correctonly if thereis a "rational"way to delimit the theoriesthat are placed
"underactive consideration."Id.
155. See id. at 42-66. For an argumentthat instrumentalrationalitydoes not exhaust all types of
rationality, see ROBERTNOZICK,THE NATUREOF RATIONALITY
133-81 (1993).
156. See Brewer,supra note 15, at 938-39.
157. See LAUDAN, supra note 143, at 23-41.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1575

methodological rules are constrained only by axiological aims, and axiological

aims are constrained by nothing in particular. Instead, Laudan argues that
constraints are multidirectional within the holistic network of aims, methods,
and beliefs. For example, factual beliefs can constrain the choice of method in
the sense that whether a given method (e.g., using double-blind experiments)
is an effective means to achieve some cognitive aim (e.g., empirical accuracy)
is itself a factual question. Some confirmation for the value of this
model-Laudan calls it "reticular"-may be found in its fertility in producing
explanations in other intellectual domains; for example such an explanation is
a prominent and important part of Dworkin's theory of "law as integrity."'58
Dworkin borrows this kind of reticulated or holistic picture to argue that what
he calls judgments of "fit" can constrain judgments of value even though the
former judgments are themselves judgments of value.159
I have argued, then, that an "enterprise"conception, properly supplemented
by Laudan's axiological model, can serve to explicate the concept of the "point
of view" in its different philosophical uses. Generalized from the particular
intellectual domain of science, the enterprise-axiological model posits that an
intellectual enterprise that produces distinctive justificatory claims may be
dissected into three separate components: factual judgments, the distinctive
methods that the enterprise uses to generate those factual judgments, and the
distinctive cognitive aims that the methods are chosen to advance and serve.
One invokes a point of view to justify some claim. To serve this justificatory
function, the point of view is assumed to be a reliable method of achieving the
(explicit or implicit) aims of some rational enterprise. Because of the holistic,
"reticulated" relation among aims, methods, and beliefs, we may understand
claims about points of view to be claims about the axiological net that is
associated with the point of view that a justificatory argument invokes.
It is time now to put this rather elaborate conception of "point of view"
to work on the principal object of my investigation: epistemic deference by
practical reasoners (specifically, legal reasoners) to expert scientists. Insofar as
legal systems and legal decisionmakers purport to base legal decisions on
justified reasoning-as virtually all legal systems claim to do-the reasoning
of authoritative legal decisionmakers in the system (like judges or juries) is
readily amenable to analysis by this axiological model sketched above. With
that model, one can compare and contrast the aims, methods, and beliefs of the
scientific expert as theoretical reasoner, with the aims, methods, and beliefs of
the judge or jury as practical reasoner. Important steps in the reasoning process
by the nonexpert practical reasoner involve an assessment by that reasoner of
the methods and results, if not also the aims, of the scientific expert, and that
assessment is (for many practical reasoners) and should be (for all such

158. See DWORKIN,supra note 3, at 176-277.

159. See RONALDDWORKIN,A MAT-ER OF PRINCIPLE167-77 (1985).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1576 The Yale Law Journal [Vol. 107: 1535

reasoners) in turn guided by certain aims that aspire to bring "intellectual due
process" into the law.

C. The Practical and Theoretical as Points of View

I am now in a position to explain my assertion that there is a distinction

between practical and theoretical reasoning that is cogent, important, and, for
my purposes, heuristically valuable.160 That distinction raises one among a
family of issues of perennial concern to moral philosophers and metaethicists,
including the relation between fact and value, between "is" and "ought,"
between descriptive and prescriptive judgments, as well as the so-called
"naturalistic fallacy."161 Nothing in my analysis of epistemic deference hangs
on whether this is in fact a tenable distinction, but it is nevertheless a very
useful heuristic. I assume with only a little argument that facts are different in
kind from values; that this is true for both moral values and the value-aims that
in large measure drive scientific inquiry; that there is an unbridgeable
inferential gap between "is" and "ought"; that any argument in which a
prescriptive proposition is putatively inferred from a set of premises containing
only descriptive propositions is either fallacious or enthymematic; and that if
it is enthymematic, then some prescriptive proposition has been omitted from
the premise set but belongs in it in a proper interpretive reconstruction of the
argument.'62(This last assumption will be prominent in my discussion below
of "practical priority.")163
Although I will not argue the fact-value issue in any detail, I will offer
some brief remarks and adduce some intuitive examples to show why I think
it tenable and useful for my particular purposes. A familiar debate about the
role that moral values play in science will help me do so. When Anglo-
American philosophy was barely emerging from the thrall of logical
positivism, Richard Rudner offered an instructive argument that the scientist

160. See supra Section II.A.

161. See, e.g., R.M. Hare,ThePromisingGame,in THEORIESOFETHICS115 (PhillippaFoot ed. 1967)
(arguingthat "ought"cannot be derivedfrom "is" withoutthe additionof a prescriptivepremise).But see
John R. Searle, How To Derive 'Ought' From 'Is,' in THEORIESOF ETHICS,supra, at 101 (arguingthat
"ought"can be derived from "is,"even withouta prescriptivepremise).One promisingrouteof inquiryis
to consider whetherclaims about what is (especially scientific claims) are on any surer epistemological
footing than are claims of morality.See, e.g., BaruchA. Brody,Intuitionsand ObjectiveMoral Knowledge,
62 MONIST446 (1979) (providingan intuitionistdefense for "objectivemoralknowledge");HilaryPutnam,
The Place of Facts in a Worldof Values,in REALISMWITHA HUMANFACE 163 (JamesConanted., 1990);
J.B. Schneewind, Moral Knowledge and Moral Principles, in REVISIONS:CHANGINGPERSPECTIVES IN
MORAL PHILOSOPHY113 (Stanley Hauerwas & Alasdair MacIntyreeds., 1983) (viewing moral and
scientific principles as enjoying equivalent rational grounding). Another important question is whether "is"
and "ought" claims are even distinguishable in kind, given the "holistic" way in which many of our beliefs
confront experience. See MORTONWHITE, WHAT IS AND WHAT OUGHT To BE DONE: AN ESSAY ON
ETHICSAND EPISTEMOLOGY (1981).
162. Thus, for example, it seems to me that Hare decisively got the better of the argument with Searle
regarding the deriviability of ought from is. Compare Hare, supra note 161, with Searle, supra note 161.
163. See infra Sections VIA-B.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1577

qua scientist must make moral value judgments."TMThe argument is

straightforward, albeit unconvincing. Its three main steps are as follows: (1) No
scientific result is certain. (2) Scientific results are measurable only on a
spectrum of epistemic confidence whose terms of epistemic evaluation vary
somewhat according to one's theory of scientific knowledge. The spectrum
might be, for example, more or less probable,'65 more or less well supported
by evidence, more or less warranted, or more or less confirmed. Per (1),
however, never does a scientific result have a probability of 1 (or the
equivalent in whatever other evaluative term is used). (3) But if (2) is true, and
if a scientist actually accepts hypotheses-as surely every scientist does-then
the scientist as such must make a moral judgment in the course of doing
science, namely the judgment that

the evidence is sufficiently strong or that the probability is sufficiently

high to warrant the acceptance of the hypothesis. Obviously our
decision regarding the evidence and respecting how strong is "strong
enough," is going to be a function of the importance, in the typically
ethical sense, of making a mistake in accepting or rejecting the
hypothesis.... How sure we need to be before we acce t a
hypothesis will depend on how serious a mistake it would be.16

Rudner buttresses his argument with three examples in which different

levels of epistemic confidence are called for in different kinds of scientific
judgment. One example concerns a scientist's hypothesis about whether a toxic
drug ingredient is present in lethal quantity in a substance to be consumed
(presumably) by humans. For this judgment, Rudner asserts, "we would require
a relatively high degree of confirmation or confidence before accepting the
hypothesis-for the consequences of making a mistake here are exceedingly
grave by our moral standards."'67But in the second example, where the
judgment is about whether, based on a sample, a particular lot of machine
stamped belt buckles is defective, we can settle for a lower level of
confidence.'68 Rudner realizes that these two examples are of a very special
kind of "scientific" judgment, namely ones in which the question of the proper
level of epistemic confidence is raised regarding "scientific inferences in
industrial quality control."'69Because his thesis is about the moral judgment
a "scientist qua scientist" must make, he then provides a third example, one

164. See RichardRudner,TheScientistQua ScientistMakes ValueJudgments,20 PHIL. SCI. 1 (1953).

165. This is Rudner'spreferreddescriptionsince he believes that "anadequaterationalreconstruction
of the procedures of science would show that every scientific inference is properly construableas a
statisticalinference."Id. at 3.
166. Id. at 2.
167. Id.
168. See id.
169. Id.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1578 The Yale Law Journal [Vol. 107: 1535

that allegedly proves his point to be "clearly quite general in application."'170

This example concerns the level of epistemic confidence on which Manhattan
Project scientists chose to rely in reaching their conclusion that no
uncontrollable pervasive chain reaction would occur when the first atomic
bomb was detonated; presumably this would be the highest level of the three
examples.'7'
Rudner's argument, and the examples he adduces to support it, fail. They
do not show that moral decisionmaking is a necessary part of scientific
decisionmaking. It is interesting to note that all three examples are ones in
which a scientist is put in the position of making some kind of practical
decision based on his theoretical occupation. That is, all examples are of the
theoretical reasoner making a practical decision based on his own theoretical
judgment. This is fairly close to the paradigm case of epistemic deference in
which I am interested; indeed, the scientist in Rudner's first example of
"industrial quality control" faces a question quite similar to that faced by the
expert witnesses in the original Daubert case172- whether Bendectin could
cause birth defects. But there is one critical difference between Rudner's
examples and my own. In my settings, the nonexpert practical reasoner must
defer epistemically to the theoretical expert to reach the practical
judgment.173
Rudner's examples do indeed call quickly to mind the moral considerations
that might in various ways impinge on a scientist's decision, but they are also
misleading for precisely that reason. Rudner adduces examples of situations in
which a scientist's decision about how to act or what to recommend does have
clear practical consequences, including moral consequences. And, of course,
on some metaethical views, moral considerations are overriding for all actions,
including the "action" of endorsing a hypothesis in testimony of various kinds
and of acting on that hypothesis by taking it to be true for the purpose of
guiding further research. Moreover, many people, including many (but by no
means all) scientists, more or less self-consciously endorse some such
metaethical view about the overriding importance of moral obligation. But

170. Id.
171. See id. at 2-3.
172. See supra note 21.
173. Perhaps Rudner's scientist should defer epistemically to a moral theorist. There is reason to
believe that moral deference takes place-children surely rely on it, and religious adherentsprobablydo
as well. Indeed,the concepts of a "moralleader"and "moralauthority"are not at all alien to the discourse
of communalmoral life. In a well-knowndissent, JudgeFranksuggestedthat the properway for the court
to go about assessing whether an immigrant had "good moral character,"as required by a federal
immigrationstatute, was to ascertain"the attitudeof our ethical leaders,"an attitudethat "would not be
too difficult to learn."Repouille v. United States, 165 F.2d 152, 154 (2d Cir. 1947) (Frank,J., dissenting).
Interestingissues remainto be analyzed regardingthe kind of epistemic deferencethat could occur
in moral settings, and whetherand when such deferenceshould occur-"should" from both epistemic and
moral points of view. As mentioned above, there appear to be cases in which courts defer to moral
philosophers"as such." In Daubert, the SupremeCourtarguablydeferredepistemically to philosophers,
albeit not moral philosophers.See supra notes 58-60 and accompanyingtext.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1998] Scientific Expert Testimony 1579

even if such a metaethical view is correct-and I doubt that it is174-the

overriding nature of moral considerations would no more show that moral
decisionmaking is a necessary part of scientific decisionmaking than would a
scientist's morally-based decision not to run a machine whose effluences he
knew would injure many people.
Nor is the scientist's decision about the level of confidence he requires in
order for him to accept a hypothesis necessarily a moral decision that is part
of science. Suppose that the best genetic research established to a confidence
level of 97% (or to a very high degree of warrant) that members of a given
ethnic group were genetically predisposed to be less intelligent than members
of other ethnic groups. As a self-consciously moral doxastic actor, a geneticist
might refuse to accept or endorse any such hypothesis unless and until the
level of confidence was greater than, say, 98% or 99%; indeed, he might even
require a confidence level that was infinitely high, in the way perhaps that a

174. The metaethicalissue Rudnerraises-whether a scientistas such mustmakemoraljudgments-is

closely analogousto one that has occupied a good deal of legal theory,namely,whethera judge "as such"
must make moraljudgments.John Lyons takes a position on this issue quite similarto Rudner's,and both
of them raise (and, to my ears, beg) the metaethical question about the overriding nature of moral
judgments. Arguing for the view that the judge as such must make moral decisions, Lyons asserts that
[p]eople who act in the nameof the law do thingswhich would requirejustificationif they were
not done in the name of the law-they use coercionand force, they kill and maim, they deprive
people of libertyand valued goods.... [Jiudicialdecisions, like other things, stand in need of
full-fledged moraljustification.
David Lyons,Substance,Process, and Outcomein ConstitutionalTheory,72 CORNELLL. REv. 745, 761-62
(1987). Basically, Lyons's contentionamountsto the claim that moralconsiderationsare overridingfor the
judge, just as Rudner'sbasic contentionis that they are overridingfor the scientist. Any point of view can
be imperialistic,demandingthat its methods, aims, and results trumpothers if and when they conflict. In
this, Rudnerand Lyons share a familiardeontic metaethicalview. ThoughI do not argue it here, my own
preferredmetaethicalview treatsmoral norms as hypothetical,not categoricalimperatives.It is a kind of
voluntarismaboutmoralobligationcogently arguedby PhillippaFoot (later,unfortunately,abandoned),and
is one that might have been better admired in Sartrianexistentialism had it been advanced in a more
disciplined manner.
An additional comment about intellectual "imperialism"is relevant. The domain of metaethics
provides one example of the fact thatphilosophical analysis is the ultimateand inevitablecourt of appeal
for disputes between and among competingpoints of view. Of course, that claim itself is advancedfrom
a philosophicalpoint of view; it's philosophy all the way up. (Ludwig Wittgensteinmay have been right
to suggest that there is no "meta-philosophy"in the sense of escaping philosophy, see LUDWIG
WiTrGENSTEIN, PHILOSOPHICAL INVESTIGATIONS ? 121 (1953), but wrong if he was also claiming that
philosophical reflection on philosophy is not possible. It is. Indeed, it is an almost inevitable, if often
implicit, part of philosophicalreflection.)What then are the distinctive aims, methods, and results of the
philosophicalpoint of view? As an axiological enterprise,philosophy'scognitive aims are not so much to
produce truthsor even correct beliefs-though truthis a regulativeideal for philosophicalanalysis. The
principal overarching cognitive aim of philosophy is to produce rationally warrantedargument and
explanation. Its methods are varied but tend to be drawn from a delimited conceptual toolkit. Those
methods deploy and are concernedwith the metaphysicalmodalities(necessity, possibility, impossibility,
contingency) and their normative kin (obligation, permission, prohibition). Philosophers deploy the
modalities to explicate "the nature"of various abstractionsand the concepts used to express them (such
as truth, meaning, knowledge, justification, justice, good, bad, evil), as well as the concepts comprising the
metaphysical and normative modalities themselves. They search for the necessary and sufficient conditions
that constitute those concepts, or some of the necessary or sufficient conditions, or try to discern why it
is not possible to offer the necessary and sufficient conditions either of some given concept or of any
concept at all. Philosophers are also centrally concerned with argument, both as an object of study and as
a method of analysis. My arguments in this Article concern several of these distinctively philosophical
enterprises, especially the last.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1580 The Yale Law Journal [Vol. 107: 1535

mother might refuse to be convinced by evidence of her son's guilt no matter

how overwhelming it might seem to others. But as a scientist, he might well
choose to require the same level of confidence for the finding about genetic
ethnic inferiority that he requires for any other scientific finding (presumably
lower than 98%, certainly lower than 100%). If he did require only the same
level of confidence, he would not be failing to serve the cognitive aims of
science-or so my intuitions run. That the epistemic demands of morality
might in such a situation (and in the situations Rudner mentions in his three
examples) be more stringent than those of science fails to show that the
scientist qua scientist is also a moralist.
Though Rudner's argument does not get him where he wants to go, his
analysis is nevertheless quite helpful in bringing out an important feature of
epistemic deference by nonexpert legal (and other practical) reasoners to
scientific (theoretical) experts. The practical point of view (legal, moral, or
prudential) consists of particular aims, methods, and judgments. As I argue in
greater detail below,'75 one of the necessary components of any practical
decision (whether or not that decision calls for epistemic deference to an
expert) is a judgment about the required level of confidence the practical
reasoner must have in the descriptive judgments he must make along the way
to reaching a practical conclusion. Rudner's examples were of theoretical
reasoners who, in Rudner's view, were inevitably driven to make practical,
moral decisions. My focus here is on practical, legal decisionmakers who in
the course of reaching a practical judgment must settle upon a required level
of confidence for those factual, theoretical judgments that are important
premises in the practical syllogism. (Again, they must settle upon the level of
confidence whether or not they solicit expert scientific testimony, but my focus
is on situations in which they do solicit it.) Like the scientists in Rudner's
examples, practical reasoners must choose among different levels of confidence
(95%, 75%, 51%, and so on). They must also choose the standard of epistemic
appraisal-"probable," "warranted,""evidentially supported,"etc.-on whose
metric the level of confidence is to be measured. Rudner's argument
acknowledges the importance of identifying these two distinct elements in
making any descriptive judgment. Recognizing both elements is a central part
of the analysis of epistemic deference that I present below.'76
In sum, although the family of distinctions between practical and
theoretical reasoning and fact and value is still much contested, these
distinctions are both useful and important for my analysis in at least two ways.
First, I will show that reasoning by a practical reasoner that involves soliciting
theoretical information from a scientific expert actually involves two distinct
lines of reasoning by the nonexpert. One line involves whatever reasoning task

175. See infra Section VI.D.

176. See infra Section VI.C.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1581

the nonexpert is pursuing that has led him to turn to the expert in the first
place (my principal concern in this Article is with legal reasoning tasks); the
other is a line of reasoning about the methods and conclusions the expert
scientist reaches in his own reasoning process. Drawing on the distinction
between practical and theoretical reasoning will be of significant heuristic
value in the discussion that follows, for it will facilitate analysis of these
converging lines of reasoning in practical epistemic deference.
As heuristically useful as the practical-theoretical distinction will be, my
analysis of epistemic deference would survive with only minor changes even
if it turned out that the distinction is not, in the end, philosophically tenable.
The object of analysis as I frame it below-deference by a nonexpert practical
reasoner to a scientific theoretical expert (what I have been referring to with
the shorthand phrase "practical epistemic deference")-could very easily be
reframed to omit references to practical and theoretical reasoning while
retaining all of the epistemological consequences of the analysis. The same
question about the cogency of epistemic deference arises when the nonexpert
epistemic "deferer" is not a practical reasoner; that question is raised, for
example, when a nonmathematician decides whether he believes that Fermat's
Last Theorem has been proved after Andrew Wiles tells him that he has
proved it, shows him the proof, and so on.'77
The people who are the epistemic "deferers" and the people who are the
epistemic "deferees" can be more "generic" on both sides of the relation "A
defers epistemically to B" than they are in the special case that is my principal
focus: practical reasoners deferring to scientific experts. On the A side, the
deferer need not be a practical reasoner. Nor need the deferee on the B side be
a scientific expert; indeed, the deferee need not be an expert at all, at least in
a common sense of that term. We will want to understand what features the
epistemic deference relation has to understand better what is special about the
relation when there is a practical (legal) reasoner on the A side who, in the
course of practical reasoning, defers to a scientific expert on the B side. For
now, I offer some observations about the general concept of epistemic
deference and then build on them to note some features of epistemic deference
in the more specific setting of practical epistemic deference.
When there are no special constraints on either A or B in the relation "A
defers epistemically to B," the epistemological issue presented is whether
testimony can be a genuine source of either knowledge or justified belief. I
discuss below a few epistemological accounts of testimony as a general source
of knowledge.'78 For the moment, I focus not on the epistemic probity of
testimony in general, but on the distinct conceptual question of what exactly
it is for one person to defer epistemically to another. Again, answering that

177. For a discussion of Wiles's proof and the publicity it occasioned, see SINGH,supra note 1.
178. See infra Section V.B.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1582 The Yale Law Journal [Vol. 107: 1535

more general question illuminates the specific analysis of practical epistemic

deference to experts.

III. "THEORETICAL"AUTHORITYAND EPISTEMICDEFERENCE

A. Believing a Person. Versus Believing a Proposition

To bring out what I think is the correct structure of general epistemic

deference I use as a foil an account offered by H.L.A. Hart.'79Hart is, of
course, chiefly concerned with the structure of practical authority, but he
argues that practical and theoretical authority (and practical and theoretical
deference) have the same underlying structure. Hart's account is importantly
inaccurate but so typically clear and insightful as to be deeply instructive even
so. Outlining the inaccuracies in his account will help me explicate precisely
the related concepts (and the relation between the concepts) of epistemic
deference and epistemic authority; to highlight the important, if obvious,
distinction between believing a person and believing a proposition; to show
that, contrary to the view Hart defends, epistemic deference and epistemic
authority are matters of degree, not all-or-nothing affairs; and to show that the
proper conceptual analysis of epistemic deference cannot by itself provide an
account of the possibility of justified epistemic deference.
According to Hart,

[T]he reason for belief constituted by a scientific authority's statement

[that some proposition is true] is in a sense peremptory since it is
accepted as a reason for belief without independent investigation or
assessment of the truth of what is stated. It is also content-
independent since its status as a reason is not dependent on the
meaning of what is asserted so long as it falls within the area of his
special expertise.
... To be an authority on some subject matter a man must in fact
have some superior knowledge, intelligence, or wisdom which makes
it reasonable to believe that what he says on that subject is more
likely to be true than the results reached by others through their
independent investigations, so that it is reasonable for them to accept
the authoritative statement without such independent investigation or
evaluation of his reasoning.'80

We should attend carefully to several points in this account of theoretical

authority.

179. See H.L.A. HART, ESSAYS ON BENTHAM: STUDIES IN JURISPRUDENCE AND POLITICAL THEORY
(1982).
180. Id. at 261-62.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1583

Hart's account seems to capture adequately a distinction that is important

for the proper analysis of epistemic deference: the distinction between
believing a proposition and believing a person. Where S is some speaker
offering testimony that p and H is a hearer of that testimony, it is the
distinction between H's believing that p and H's believing S that p. The
distinction has long been remarked by philosophers. Thomas Hobbes, for
example, noted: "To have faith in, or trust to, or believe a man, signify the
same thing; namely, an opinion of the veracity of the man: but to believe what
is said, signifieth only an opinion of the truth of the saying."'8' Several
recent works in the epistemology of testimony also make, and rightly insist on,
the importance of this distinction. G.E.M. Anscombe argues that "[o]ne might
think at first blush that to believe another is simply to believe what he says,
or believe what he says is true. But that is not so, for one may already believe
the thing he says."'82 She adduces the example of a person who reports
believing that a man was dying "'[b]ecause the doctor told me .... I had no
opinion of my own-I just believed the doctor,"",183 and argues that the
example "brings out how believing x that p involves relying on x for it that
p. "184

Tony Coady makes this distinction a prominent part of both his general
account of testimony, and his explication of the pragmatic conventions that
govern the "speech act of testifying."'85 According to Coady, there are three
individually necessary and jointly sufficient conditions under which speaker S's
statement that p constitutes the speech act of testifying:

A speaker S testifies by making some statement p if and only if: (1)

His stating that p is evidence that p and is offered as evidence that p;
(2) S has the relevant competence, authority, or credentials to state
truly that p; (3) S's statement that p is relevant to some disputed or
unresolved question (which may, or may not be, p) and is directed to
those who are in need of evidence on the matter.'86

On Coady's account, in testimonial settings, it is the fact that S asserts p that

is evidence for p's truth. That is, what the hearer of S's testimony believes is
S's assertion that p, not simply p itself (though, of course, when the hearer

181. THOMAS HOBBES, LEVIATHAN (1651), reprinted in 3 THE ENGLISH WORKS OF THOMAS HOBBES
OF MALMESBURY 1, 54 (Sir William Molesworthed., London,John Bohn 1839).
182. G.E.M. Anscombe, WhatIs It To Believe Someone?, in RATIONALITY AND RELIGIOUSBELIEF
141, 144 (C.F. Delaney ed., 1979).
183. Id. at 145.
184. Id. (emphasis added).
185. C.A.J. COADY, TESTIMONY:A PHILOSOPHICAL STUDY (1992).
186. Id. at 42.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1584 The Yale Law Journal [Vol. 107: 1535

does believe S's assertion that p, the hearer also believes p'87), and S's
assertion that p is distinct from p itself with regard to the probity of p.'88
There are various ways to establish the cogency of the distinction between
believing that p and believing S that p. One is to observe that in most cases in
which a speaker S testifies that p, p itself is (at the very least) unlikely to be
evidence of its own truth, whereas in many cases (when Coady's conditions
(2) and (3) are also satisfied, for example) S's testimony that p is evidence that
p. There are distinct reasons for this difference. One is the substantial difficulty
of identifying propositions that are self-evident, a difficulty that I shall not
rehearse here. Another reason, partly linguistic, partly epistemic, is as follows.
In thoroughly Gricean spirit, Coady's conditions (2) and (3) suggest that
testimony is not properly cooperative unless the witness has good reasons for
asserting p'89 and the testimony is relevant to and informative regarding
some inquiry or interest of the hearer.'90Thus, if we assume-as I do-that
Grice is correct in his view that speakers and hearers do by and large obey the
cooperative norms he identifies, the hearer should have some significant doubt
about the truth of p in the normal case of testimony (in which the speaker is
not, for example, trying to generate a conversational implicature). With less of
the supporting machinery, Hobbes put the same point more succinctly: "[N]o
man is a witness to him that already believeth, and therefore needs no witness;
but to him that deny or doubt, or have not heard it."'9' It is doubt about the
subject matter of proffered testimony that enables that testimony to be
cooperative-that is, doubt about p looks like a necessary condition (albeit a
defeasible one) of the cooperativeness of testimony that p. When that
necessary condition obtains, p should not constitute its own evidence.'92
Thus, believing that p is very different from believing a person's testimony
that p.

187. Again, Hobbes was on targethere:

When a man's discoursebeginneth... at some saying of another,of whose ability to know the
truth,and of whose honesty in not deceiving, he doubtethnot; and then the discourseis not so
much concerningthe thing, as the person;and the resolutionis called BELIEF, and FAITH:faith,
in the man; belief, both of the man, and of the truthof what he says.
HOBBES, supra note 181, reprinted in 3 THE ENGLISHWORKS OF THOMASHOBBESOF MALMESBURY,
supra note 181, at 53-54.
188. As Coady explains his view:
[T]estimonyis the evidence of persons[,] and [condition(2)] shows in a schematic way what
it is about persons that makes this sort of evidence special.... [In the case of testimony] we
are not just believing thatp because of somethingor otheraboutthe witness's utterancebut we
are believing the witness.
COADY,supra note 185, at 46.
189. ConsiderPaul Grice's maxim of quality:"1. Do not say what you believe to be false. 2. Do not
say that for which you lack adequateevidence."PAUL GRICE,STUDIESIN THEWAY OF WORDS27 (1989).
190. CompareGrice's maxims of quantityand relation.See id. at 26-27.
191. HOBBES, supra note 181, reprinted in 3 THE ENGLISH WORKS OF THOMAS HOBBES OF
MALMESBURY, supra note 181, at 496.
192. A very helpfuldiscussion with CatherineElgin alertedme to a relatedpoint, which I here extend
to the question of whethera testifiercould cooperativelytestify regardingsomethingself-evident.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1585

Though he does not discuss the point directly, Hart's account of epistemic
deference appears to recognize this distinction. It seems fairly presupposed by
his assertion that "the reason for belief constituted by a scientific authority's
statement [that some proposition p is true] is in a sense peremptory."'93Here
it is not just the proposition that constitutes the peremptory reason, but rather
the scientific authority's statement of it that does. Similarly, Hart maintains
that when the witness satisfies certain conditions of competence, it becomes
reasonable for a hearer to "accept the authoritative statement without...
independent investigation or evaluation of [the witness's] reasoning."'194 Here
again, though not quite explicitly, Hart seems to recognize the distinction
between accepting the statement that p and accepting the authoritative
statement that p, the latter of which plausibly carries with it the idea that p is
believed because it issues from this authoritative person.

B. Actual Versus Putative Authoritativeness

I turn to another feature of Hart's account of epistemic authority, its

requirement that a person have actual competence rather than merely putative
competence to be considered an authority. Recall that, on Hart's account, a
putative epistemic authority, A, is not an actual epistemic authority unless A
"in fact" possesses sufficiently superior wisdom (etc.) making it reasonable for
a deferring nonexpert, B, to believe what A says without any independent
investigation by B.'95 Hart quite clearly holds this position. 196 For Hart, the
judgment that authority is warranted (warranted by the fact of superior
knowledge) is built into the concept of authority itself.

193. HART, supra note 179, at 261.

194. Id. at 262.
195. See id.; see also supra text accompanying note 180. Other theorists of practical and theoretical
authority differ as to whether actual possession of superior knowledge (and related traits or information)
is a necessary condition of epistemic authority. Heidi Hurd, for example, offers an account of epistemic
authority that seems similar to Hart's on this point, asserting that "[t]he advice of a theoretical authority
constitutes evidence at all only because that authority is more motivated to discover the truth, or is in
possession of more information, or has superior inference-drawing abilities." Heidi M. Hurd, Challenging
Authority, 100 YALE L.J. 1611, 1669 (1991). Raz's position on this point is harder to assess. On the one
hand, he seems sympathetic with Hart's view in using the "dependence thesis" as an account of both
theoretical and practical authority, according to which "all authoritative directives should be based on
reasons which already independentlyapply to the subjectsof the directivesand are relevantto their action
in the circumstancescovered by the directive."JOSEPHRAZ, THE MORALITYOF FREEDOM47 (1986). On
the other hand, Raz also expressly denies that "authoritative determinations are binding only if they
correctly reflect the reasons on which they depend." Id. (emphasis added). "On the contrary," he argues,
"there is no point in having authorities unless their determinations are binding even if mistaken ...." Id.
Raz suggests that practical authorities and theoretic authorities "share the same basic structure" that is
expressed by the dependence thesis. Id. at 53. If the analogy extends to legal authorities specifically, Raz's
position is probably this: Just as a legal authority need not necessarily be legitimate to be a legal authority,
but rather need only claim to be so, so an epistemic authority need not necessarily have the requisite
epistemic competence in the subject matter of his putative expertise, but rather need only claim to have it,
to be an epistemic authority.
196. See HART, supra note 179, at 262.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1586 The Yale Law Journal [Vol. 107: 1535

In my view, an account of epistemic authority should be neutral as to

whether in any given instance the granting of that authority is warranted.
Hart's conception of authority makes it difficult for an analyst to describe what
is surely a common phenomenon, that of a person A being treated as an
epistemic authority by some group of persons B even though a third-party
analyst, C, thinks (contrary to B's judgment) that A does not actually have
superior intelligence, wisdom, etc. Accordingly, I would modify this definition
in the following way (it will be modified still more in subsequent discussion):
For A to be an epistemic authority for B on some subject matter, B must
believe that A has some superior knowledge, intelligence, or wisdom which
makes it reasonable to believe that what A says on that subject is more likely
to be true than the results reached by B through B's independent investigations.

C. The Concept of Epistemic Deference

Hart's account seems importantly correct insofar as it identifies content-

independence and peremptoriness as chief features of epistemic authority (i.e.,
expertise). If these conditions are treated as absolute, however, the account
deprives us of a useful analytical tool in examining the practice of epistemic
authority. That is, Hart's analysis of the concept misses an important part of
the phenomenon if it endorses the following claim: It is inherent in the
structure of the concept of epistemic authority that, in order for A to be an
epistemic authority for B, B must treat A's judgments (within the sphere of A's
authority, as understood by B) as absolutely content-independent and
peremptory.
As I have said, the basic intuition about the role of content-independence
and peremptoriness seems correct and important. Surely, part of what it is for
A to be an epistemic authority for B is for B to treat A's judgment as providing
a significantly peremptory and content-independent reason for believing what
A says. Thus, a non-physicist does not and should not check to see what
physics proposition the physicist endorses in order then to decide whether to
defer to that physicist about the truth of the proposition. To do so would be
precisely to deny the physicist's expertise and epistemic authority. But there
is also good reason to believe that epistemic deference, and, concomitantly,
epistemic authority, is a matter of degree, not an all-or-nothing relationship.
For one thing, Hart himself acknowledges that, when B treats A as an
epistemic authority, B's deference extends only so far as B recognizes A to be
speaking within the subject area of A's expertise.'9' Thus, even in
paradigmatic cases of epistemic deference, the nonexpert must police the
epistemic boundaries between assertions by A that are within what B

197. See id. at 261 ("Itis also content-independentsince its statusas a reasonis not dependenton the
meaning of what is assertedso long as it falls within the area of his special expertise.").

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1998] Scientific Expert Testimony 1587

recognizes to be the zone of A's expertise and assertions by A that are not
within that zone. These borderlines will inevitably be fuzzy. For example,
where exactly does the epistemic authority of a physicist end when he is
testifying to the nonexpert about the advisability of nuclear energy? Not
exactly anywhere. Even if such an expert is testifying as an "instrumentally
rational" expert, one who is using his expert knowledge to advise the
nonexpert about how best to achieve goals the nonexpert has chosen, the
nonexpert must be ever vigilant to keep the expert within his proper epistemic
domain. The price of rational deference is eternal vigilance.'98
Hart's all-or-nothing criterion is also too strong in that it would condemn
the concept of epistemic authority to overlook a flourishing practice of
epistemic deference among epistemic equals. This practice abounds, for
example, among scientists who work together on large research projects and
divide epistemic labor among themselves for the sake of efficiency, even when
they are epistemically capable of doing the work themselves. A related
phenomenon is the practice in the scientific community of scientists relying on
other scientists' results cumulatively. Cumulative reliance is at the very least
a familiar phenomenon within the scientific community, and some have even
argued that it is a necessary condition of scientificity.'99 Cumulative reliance
of this sort constitutes a type of epistemic deference that ought not to be
defined away.
Hart's account also seems unable to capture the phenomenon of
"persuasive authority"among what we might call epistemic "near-equals."This
is a common phenomenon in legal reasoning. A "persuasive authority"
functions to some degree as an epistemic authority even for decisionmakers
who are themselves substantially competent in the areas the persuasive
authority addresses. An eminent treatise writer might give a judge compelling
reason to believe that the law is as the writer claims. Such a writer's work will
often function as an epistemic authority for a judge, but it is not the case that
this work is "accepted as a reason for belief without independent investigation

198. Otherexamples come to mind. Whereexactly does the epistemic authorityof a psychologistend
when he is testifying about the degree to which a victim of battered-women'ssyndromewas responsible
for her actions?Where exactly does the epistemic authorityof a medical doctorend when she is advising
nonphysiciansat the Immigrationand NaturalizationService regardingwhethera homosexualalien should
be excluded from the United States under a statute that forbids immigrationto persons who have
"psychopathicpersonality,sexual deviation,or mentaldefect"?Hill v. INS, 714 F.2d 1470, 1475 (9th Cir.
1983) (quoting 8 U.S.C. ? 1182(a)(4) (1970)). These and many other adducibleexamples of the vague
bordersthat demarcatethe zone of the expert's special competence reveal that the content-independence
of a nonexpert'sreasonfor deferringcannotbe absolute.See id. ("Itwould thus violate Congress'direction
to allow INS officers who are not medicallytrainedto determinepsychopathicpersonality,sexual deviation,
or mental defect by interrogation.... Congress intendedthe determinationto be made solely by trained
physicians.").
199. Anthony Kenny suggests in his discussion of expert testimony in courts that the possibility of
cumulativereliancewithina putativescientificdisciplineis indeeda necessaryconditionof the discipline's
being a science. On Kenny's view, "thoughany expertmustbe able to repeatthe resultsof others he does
not have to: he can build on the foundationsthatothershave built."Kenny,supra note 64, at 50 (emphasis
added).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1588 The Yale Law Journal [Vol. 107: 1535

or assessment of the truth of what is stated."200 Rather, it functions as

persuasive, though not dispositive, advice. This authority's endorsement of the
propositions can function as a signal that those propositions ought to be taken
very seriously, even over and above the judge's own first impression (if he has
one). The court may thus accord the authority's view some prima facie weight
while also making an "independent investigation or assessment of the truth of
what is stated.",201
Accordingly, I propose an additional modification of Hart's account:

For A to be an epistemic authority for B on some subject matter, B

must judge that A has some sufficient knowledge, intelligence, or
wisdom which makes it reasonable to believe either that what A says
on that subject is more likely to be true than the results reached by B
through B's independent investigations, or is no less likely to be true
than the results that would be reached by B through B's independent
investigations.202

Having recognized that epistemic deference and epistemic authority are matters
of degree, I should also observe that by no means every instance of epistemic
deference raises the specific philosophical problem I have set out to explore.
That problem pertains only to the deference that a nonexpert gives to an
expert.
Although Hart clearly believes that justified epistemic deference by
nonexperts to experts ("theoretical authorities," in his terms)203is possible,
his account of epistemic authority and deference does not in any way explain
how it is possible. What it says, slightly rephrased, is that A is an expert if,
vis-a-vis the nonexpert, B, A has superior knowledge, intelligence, or wisdom
that makes it reasonable for B to believe that what A says on that subject is

200. Id.
201. Id.
202. Sometimes it will be vacuously true that B treatsA as an epistemic authority.For example, in
situationsin which B has not made any independentinvestigationon a given subject and A has, B could
judge thatA "has sufficient knowledge, intelligence,or wisdom which makes it reasonableto believe that
whatA says on that subjectis more likely to be truethanthe resultsreachedby B throughB's independent
investigations"-simply by virtue of the fact thatB made no independentinvestigationand plannednever
to make such investigation. Such a judgment by B may be vacuously true, but it would well be true
nevertheless. Despite the vacuity, it is useful to illustratethis possibility with a hypotheticalexample.
Suppose B, an experiencednutritionist,is hiredas a consultantfor C. A is a college studentwho works for
B. B has decided it would be useful, albeit not that importantin the long run, to determinehow many
calories, on an average daily basis, C consumedover a specified time. Both A and B have access to a list
of all the foods and quantitiesof food C consumedduringthattime. It could well be the case thatB treated
A as an epistemic authorityon the questionof C's daily averagecalorie consumption,underthe following
conditions:(1) A took the time to apply his perceptualand inferentialfaculties to the problemwhile B did
not; (2) B did not think it would be worthB's time to do so; even though (3) B also believed that had B
taken the time to do the calculation, he would have produced a more accurate one. (Thanks to Kent
Greenawaltfor useful discussion on this point.)
203. HART,supra note 179, at 261-62. I agree with Raz: "Nowadaysit is not the fashion to talk of
[theoretical]authorities.... Insteadwe have experts."RAZ,supra note 195, at 52.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1589

more likely to be true than the results reached by B through his independent
investigations, so that it is reasonable for B to accept the authoritative
statement without such independent investigation or evaluation of A's
reasoning. Nothing in this definition says what criteria nonexperts are to use
to identify these epistemically superior beings, or whether, being nonexperts,
they can deploy any such criteria in an epistemically warranted manner.2'04
Thus, if Hart is correct in maintaining that his analysis provides an accurate
account of theoretical authority, we may conclude that such an account does
not by itself provide a philosophical explanation of how justified deference to
experts is possible. I am inclined to believe that, with suitably important
modifications, it is an accurate account.205
Building on the foregoing analysis of the structure of epistemic deference,
I offer the following definitions. Although I offer them as stipulated
definitions, I also aspire to have them capture, refer to, and felicitously
describe actual linguistic practice.
An expert is a person who has or is regarded as having specialized training
that yields sufficient epistemic competence to understand the aims, methods,
and results of an expert discipline.206An expert discipline is a discipline that
in fact requires specialized training in order for a person to attain sufficient
epistemic competence to understand its aims and methods, and to be able
critically to deploy those methods, in service of these aims, to produce the
judgments that issue from its distinctive point of view.207A nonexpert is a
person who does not in fact have the specialized training required to yield
sufficient epistemic competence to understand the aims, methods, and
judgments of an expert discipline, or to be able critically to deploy those
methods, in service of the discipline's aims, to produce the judgments that
issue from the discipline's distinctive point of view.
Expert and theoretical authority are different terms for the same kind of
person or group. In accord with the definition of 'expert' offered above, a
'theoretical authority' is a person or group of persons who has or is regarded

204. Note also that Hart's account does not make knowledge a necessary condition of epistemic
authority, only
superior knowledge, intelligence, or wisdom which makes it reasonable to believe that what he
says on that subject is more likely to be true than the results reached by others through their
independent investigations, so that it is reasonable for them to accept the authoritative statement
without such independent investigation or evaluation of his reasoning.
HART, supra note 179, at 262 (emphasis added).
205. Hart offers this as his account of both practical and theoretical authority, and shares with Raz the
basic explanation of the structure of authority.
206. Though I do not try to say it here, more could be said about what constitutes the kind of
"specialized training" that can yield epistemic competence in an expert discipline. For example, certain
kinds of experience, if properly reflected upon, might well provide the requisite "training," so that a
religious mystic might well be an expert on the beliefs and requirements of his religion even without having
what would conventionally be called "training." (Thanks to Kent Greenawalt for helpful discussion on this
point.)
207. See supra Section II.B (discussing axiology and point of view).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1590 The Yale Law Journal [Vol. 107: 1535

as having the competence to render judgments from a given theoretical point

of view. To adopt as an epistemic rule "Whatever the Pope says is true" is to
treat the Pope as a theoretical authority who is competent to pronounce what
is so from the "Catholic point of view."208

IV. EXPERTISE:CONCEPTSAND BASIC PROBLEMSFOR THE NONEXPERT

A. Knowledge, Warranted Beliefs, and Degrees of Epistemic Competence

In the previous section, I defined expert (and theoretical authority) in terms

of epistemic competence, which in turn I described using the term
'understanding.' Some explanation of the concepts of epistemic and
understanding are now due. The distinctive mark of the epistemic is the
concern with warranted belief. That is, the epistemic point of view is the view
of a reasoner whose overriding cognitive goal209 is to acquire warranted
beliefs. 'Warrantedbelief' here is a placeholder. Different epistemologies fill
in this place in different ways.210 Some declare that truth is the sole
overriding cognitive goal for a properly epistemic point of view,2"' while
others allow a broader range of cognitive aims to play a role, even when they

208. This particulartheoreticalpoint of view serves also as a practical one. Indeed, it may be a
distinctive feature of religious points of view that they tend to serve both functions. According to some
realist theoriesof moraljudgment,the moralpoint of view is also simultaneouslytheoreticaland practical.
Michael Moore construesRaz's conceptionof practicalauthorityto have this kind of realist dimension:
Why isn't practical authorityalways irrationaland thus always illegitimate in its apparent
requirementthat we act against(or at least not judge for ourselves)what the balanceof reasons
indicates we should do? Raz's answer is very attractive,given in terms of what he calls the
"service conception"of authority.As the name suggests, accordingto the service conception,
authoritiesare legitimate to the extent that they serve us. Authoritiesserve us wheneverthey
promulgatedirectives which, if we follow them, producebehaviorby us that betterconforms
to what we ought to do anyway (independentlyof any authoritativedirective)thanthe behavior
which would result if we calculateddirectly what we ought to do. This is an attractivenotion
of legitimate authority because it meets the irrationalityobjection head-on: authority is
legitimateonly when it is rationalnot to act on those reasonsthat would otherwise(withoutthe
authority)make it rationalfor us to do somethingelse.
The serviceconceptionof legitimatepracticalauthoritysoftensthe line betweentheoretical
and practicalauthority.If one grants that there is such a thing as knowledge about what, all
things considered,one ought to do, then anotherperson is a practicalauthorityfor us usually
when he is also a theoreticalauthority(aboutone area of knowledge,namely,about where the
balance of our reasons for action comes out). It is this greaterknowledge of the reasons that
antecedentlybind us that often gives someone legitimateauthorityover what we should do.
Michael S. Moore,Authority,Law,and RazianReasons,62 S. CAL. L. REv. 827, 830-31 (1989) (footnotes
omitted).
209. See supra Section II.B (discussing axiology and point of view).
210. See supra Section II.C (discussing Rudner).
211. Bonjour, for example, appears to emphasize truth in this way:
The distinguishing characteristic of epistemic justification is . .. its essential or internal relation
to the cognitive goal of truth. It follows that one's cognitive endeavors are epistemically
justified only if and to the extent that they are aimed at this goal, which means very roughly
that one accepts all and only those beliefs which one has good reason to think are true.
BONJOUR,supra note 142, at 8.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1591

are not valued as instrumental means to achieving the end of discovering

truth.212In identifying the epistemic point of view with the cognitive goal of
achieving warranted beliefs, I offer no strong commitment to any particular
epistemic axiology, although it does seem that exclusively "vericentric"
epistemologies (those that treat discovery of truth as the overriding epistemic
goal) run a significant risk of misdescribing the epistemic structure of many
epistemic practices, including the practice of science.
The concept of understanding should be an important part of any full
epistemological theory. In contemporary epistemology, it has received
relatively little express treatment compared to headliner concepts like
knowledge and justification, and the treatments it has received have tended to
focus specifically on linguistic understanding. Miles Burnyeat has begun a
promising, more general line of inquiry into understanding, one very much
consonant with the approach I am taking in this Article.213It comes in his
discussion of the Platonic and Aristotelian conception of the relation between
knowledge on the one hand, and synoptic, explanatory understanding on the
other:

The important difference between knowledge and understanding is

that knowledge can be piecemeal, can grasp isolated truths one by
one, whereas understanding always involves seeing connections and
relations between the items known. "The only part of modem physics
I understand is the formula 'E = mc2"' is nonsense. "The only part of
modem physics I know is the formula 'E = mc2'" is merely sad.214

Burnyeat speaks of the philosopher who, like Plato, "wants ... to assimilate
knowledge to rational understanding."21 5 Though assimilating knowledge to
understanding or, perhaps even better yet, replacing the concept of knowledge
as the focal point of epistemology with the concept of understanding, has its
attractions,216 I have no ambition in the present work to take on so large a task.
My theoretical needs are far more modest, attempting to explicate only a notion
of epistemic competence that captures and helps to explain the nature of the
cognitive capacity involved in possessing scientific expertise. For that purpose,
the notion of understanding as the possession of a widening, explanatory,217

212. See, e.g., CATHERINEZ. ELGIN,CONSIDEREDJUDGMENT(1996).

Soc'Y
213. See M.F. Burnyeat,Wittgensteinand AugustineDe Magistro, 61 PROC.ARISTOTELIAN
SUPPLEMENT1, 20 (1987).
214. Id.
215. Id.
216. Catherine Elgin is making a serious inroad on this project. See ELGIN,supra note 212.
217. Explanation is important to the conception of understanding I endorse here. Again, Burnyeat puts
it well:
Plato, like Aristotle, makes it a condition on knowing or understanding that p that one grasps
the explanation of p. This of course involves seeing the connection between p and a whole lot
of other propositions, but it is not mere connectedness so much as explanatory connectedness
that counts, and it is by way of this thought that Plato and Aristotle reach the conclusion that

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1592 The Yale Law Journal [Vol. 107: 1535

synoptic grasp serves quite well.218

Recall the discussion of point of view and axiology, in which I extended
Larry Laudan's model of rationality into a general account of rational
enterprises and their distinctive aims, methods, and factual judgments.219
Laudan calls his model of scientific justification "reticular" to mark the way
in which the aims, methods, and beliefs of science are mutually supporting and
explaining, and how each type (aim, method, factual judgment) can occasion
a change in each of the other types. A chosen cognitive aim, for example, can
change or dictate choice of method-as when a psychologist chooses the
double-blind experiment method-because she believes it will serve the
cognitive aim of truth in psychological experiment. Note that this example also
illustrates the way in which a factual judgment helps guide the choice of
method-for it is a factual judgment in the efficacy of the double-blind
experiment (and the substantially lesser efficacy of single-blind experiment)
that leads her to choose that as a method of producing truths. Similarly, a
factual judgment about the realizability of an aim can compel a change in the
aim-for it makes little sense to pursue an aim that one decides, as a matter
of factual judgment, is unachievable. Each "node" of the net can have a
justificatory and explanatory impact on each of the others. This model of
justification, suitably extended from science to intellectual discipline, is a
holistic account that emphasizes the centrality of reflective adjustment in
epistemic justification, giving it much the same prominence as in the account
of exemplary argument I have offered elsewhere.220 And it is a model that
gives content to the kind of latticework of explanatory relations present in the
understanding of an expert discipline. To have epistemic competence in such
a discipline is, I suggest, to be capable of grasping and manipulating this kind
of reticular structure of aims, methods, and factual judgments in an expert

knowledge in the full sense, i.e. understanding,requiresthe synoptic grasp of a whole field.
Burnyeat,supra note 213, at 21.
218. The view of understandingas possession of a synoptic grasp of explanatoryrelations is not
uncommonamong philosopherswho explicitly discuss the concept. For example, Neil Cooperargues that
"understandingis concernedwith relationsand connections"and that "[i]t is possible to have knowledge
of a bitty or superficial kind, while we only have understandingwhen we relate or connect bits of
knowledge with other bits in a more or less coherent whole." Neil Cooper, Understanding,68 PROC.
ARISTOTELIAN Soc'Y SUPPLEMENT 1, 3-4 (1994). To like effect is CatherineElgin's moreambitiousproject
of articulatingan understanding-oriented, ratherthan a knowledge-oriented,epistemology:
'Understanding'is a bettertermfor the epistemicachievementthatconcernsus here. Not being
restrictedto facts, understandingis far more comprehensivethan knowledgeever hoped to be.
We understandrules and reasons,actions andpassions,objectionsand obstacles,techniquesand
tools, forms and functions and fictions, as well as facts. We also understand pictures, words,
equations, and patterns.Ordinarilythese are not isolated accomplishments;they coalesce into
an understanding of a subject, discipline, or field of study.... Understanding a particular fact
or finding, concept or value, technique or law is largely a matter of knowing where it fits and
how it functions in a matrix of commitments.
ELGIN, supra note 212, at 123.
219. See supra notes 143-158 and accompanying text.
220. See Brewer, supra note 15, at 962-78.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1593

discipline. It is precisely the lack of this kind of understanding in nonexpert

legal reasoners that casts doubt on their capacity to rely legitimately on expert
scientific testimony in reaching practical decisions.
One more point is important for my later assessments of cases like
Daubert, and for the overall conclusion of this Article. Epistemic competence
in an expert discipline comes in degrees; it is not an all-or-nothing "switch."
This is perhaps not surprising. Is it not a familiar fact that some
mathematicians, logicians, physicists, economists, geneticists, and so forth are
more skilled at grasping and manipulating the aims, methods, and factual
judgments of their respective expert disciplines than are other experts in the
same disciplines? Surely Isaac Newton was a more epistemically competent
physicist than Isaac Asimov. By the same token, we should recognize that
there is no bright line separating expertise from nonexpertise-just as there is
no bright time line or light line separating night from day, even though there
is clearly a difference between night and day. Not all experts are equally
epistemically competent in their disciplines, nor are all nonexperts equally
incompetent with regard to a given expert discipline.

B. Practical Epistemic Deference and Theoretical Judgment

As I have noted above, I use "practical epistemic deference" as an

abbreviation for deference by a nonexpert practical reasoner, like a
scientifically untrained judge or jury, to a scientific theoretical expert. This is
the reasoning process that is my central concern. Of course, not every instance
of epistemic deference by a practical reasoner to a theoretical expert is
deference by a nonexpert practical reasoner; an epistemically competent
practical reasoner could defer to an epistemic equal or near equal.221 I restrict
"practical epistemic" deference to nonexpert practical reasoners only for the
sake of abbreviation.
A "theoretical judgment" is a judgment about what one ought to believe
from an epistemic point of view. Although theoretical judgments are a central
part of scientific inquiry, not only scientists make them. Religious beliefs and
pseudoscientific judgments, such as those of astrology or necromancy, as well
as judgments in the literary and plastic arts, are also "theoretical" judgments
in this sense.222

C. The Nonexpert's Selection and Competition Problems

The nonexpert faces at least four distinct problems, which I will call
"selection problems." To explain them, I begin with a simple (and vague, but

221. See supra Section III.B.

222. See NELSON GOODMAN, WAYS OF WORLDMAKING 1-7 (1978).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1594 The Yale Law Journal [Vol. 107: 1535

nonetheless sufficient) definition. Call HL the overall hypothesis of a case

presented to a practical reasoning authority (like a judge or jury). HLmight be,
for example, the prosecution's claim that Jones committed the murder, or the
plaintiff's claim that Smith breached the contract. (HL will usually be an
"ultimate" issue in the case, and a mixed question of law and fact.) Call each
Hi an individual hypothesis whose conjunction with all the other Hi implies H,.
(H, need not also imply the Hi.) An evidentiary proposition (which itself can
be logically simple or complex), call it e, is objective evidence for some Hi,
just in case Hi is better warranted given the truth of e than it is given the
falsity of e. An evidentiary proposition e is rationally pertinent to some Hijust
when it constitutes objective evidence of that Hi. (That is, 'objective evidence
for' and 'rationally pertinent to' are synonyms.) An evidentiary proposition e
is rationally pertinent to the overall hypothesis of a case, Hc, just in case e is
rationally pertinent to some Hi (which, as defined, is in the set of propositions
that imply H). This definition is a slight generalization and amalgamation of
the concepts of relevance and materiality found in the common law of
evidence223 as well as a basic definition of objective evidence found in the
epistemological literature.224A virtue of the definition's vagueness is that it
remains neutral among various standards of epistemic appraisal and attendant
levels of confidence.
When a nonexpert judge or judge andjury (who may divide decisionmaking
labor in the familiar ways, relying on judgments of admissibility, relevance and
materiality, sufficiency, and weight) must decide whether to consult a putative
scientific expert in the course of deliberating about Hc, that nonexpert faces four
"selection problems": (1) determining which of the intellectual enterprises that
might yield expert testimony is a science; (2) determining who is a scientist
capable of using her science in a manner that satisfies the standardof epistemic
appraisal and the attendant level of confidence that the practical reasoner has
established; (3) determining which of the intellectual enterprises that might
yield expert testimony is a science that is rationally pertinent to the case (that
is, to H); (4) in cases in which there is significant doubt occasioned by task (3),
determining who is capable of answering (3) in a way that can identify an
expert scientific discipline capable of satisfying the chosen standard of
epistemic appraisal and the attendant level of confidence.
"Competition" in expert testimony occurs when two experts testify to
evidentiary propositions that are either contrary or contradictory.225The fact

223. See supra note 19 and accompanyingtext.

224. See, e.g., Peter Achinstein,Conceptsof Evidence, 87 MIND 22 (1978). As Achinsteindiscusses,
this conception of evidence is "objective"in that e's being evidence for Hi depends neither on anyone
believing e or Hi nor on anythingabout their relation.See id. at 23.
225. Logicallyandepistemologicallyweakerformsof competition,suchas "incoherence,"arepossible,
but the more easily articulableand understoodforms of contradictionand competitionwill suffice for my
analysis.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1595

that much scientific expert testimony is competitive in this way presents a

particular problem and puzzle for systems in which nonexperts make decisions
that rely on expert testimony. When experts disagree about the truth of some
evidentiary proposition e, the nonexpert must decide whom to believe on the
scientific issue. But, ex hypothesi, the nonexpert does not have sufficient
competence in the expert discipline to be able to make the choice on
substantive grounds, so how can the nonexpert make that choice? If we assume
honesty on the part of each expert, this can seem especially puzzling in that
it may look like we are expecting greater ability to discern the scientific truth
from the nonexpert than we are from the expert. This is so because the rules
of evidence that govern admissibility of expert testimony (i.e., what testimony
the nonexpert jury or judge is allowed to hear) do not presuppose that all the
experts are testifying to the truth. This we know in part from the basic fact that
judges "qualify" experts who the judges know will testify to contrary or
contradictory propositions-even when those judges are also fully aware that
two evidentiary propositions that are mutually contrary or contradictory cannot
both be true. But if a judge does not expect that every scientific expert is
testifying to the truth (and thus, on the assumption that the expert is testifying
honestly, the judge is not expecting every expert to know the truth either), then
is the judge expecting the nonexpert judge or jury actually to do better than the
expert at discerning the truth? It might seem so, because obviously each expert
has an opportunity to hear what the opposing expert will say, and thus has as
much opportunity to revise his own contrary beliefs as the jury has to form its
beliefs.
There are intra- and extra-disciplinary versions of the problem of
competition. The intra-disciplinary version arises when two experts within the
same field, e.g., epidemiology, testify in contrary or contradictory ways about
some issue. The extra-disciplinary version arises when experts from different
fields (e.g., philosophy and psychiatry) testify in contrary or contradictory
ways about some issue. (I assume that fields, roughly academic fields, have
real epistemic significance, and are not merely matters of university
administration.) There are also what I shall call problems of actual competition
and implied competition. Actual competition occurs when the testimony of two
or more competing experts is actually admitted at trial. Implied competition
occurs when there is an expert opinion on a particular subject or topic that
exists "extra-camerally," and that, if it had been admitted, would have created
a problem of actual competition. One can also imagine a kind of diachronic
competition in expert testimony, in which claims that experts support at one
time become challenged by experts at a later time.226The basic problem
about competition is how a nonexpert faced with competing expert testimony
about some evidentiary proposition can decide which of the competing experts

226. I briefly pointed to this kind of competitionin the discussion of Brown.See supra Section I.C.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1596 The Yale Law Journal [Vol. 107: 1535

(intra-disciplinary or extra-disciplinary, actual or implied) to believe without

being able to assess the substantive merits of the competing arguments they
offer to support their evidentiary judgments.

V. KNOWLEDGEVERSUS JUSTIFIEDBELIEF: WHAT Do

NONEXPERTS WANT FROM EXPERTS?

A. What the Law Desires: Acquiring Justified Beliefs from Experts

One can assess a given state of mind as simple belief, or as knowledge, or

as a belief that is well grounded or rational or justified-and so on. These
categories might be arrayed on a spectrum (for example, with simple belief at
one end, knowledge at the other, and some of the additional terms of epistemic
evaluation somewhere between), or might be explicated as more or less
separate categories, each with its own criteria. It is the category of knowledge
that has dominated the attention of modern epistemologists. In recent decades,
this attention has often taken the form of considering the modern "classical"
criteria of knowledge as justified true belief, and looking for additional or
different criteria of the concept that are capable of handling Edmund Gettier
problems and the like.227More recently, epistemologists have offered theories
containing robust explications of other terms of epistemic evaluation (though
some of these have in turn been offered principally as accounts of some of the
classic tripartite criteria of knowledge), such as justification, coherence,
reliability, and evidence.
I am concerned in this Article with the epistemic competence of nonexpert
judges and juries when, in the course of making a legal judgment, they assess
putatively relevant and material scientific information. But in order to assess
their epistemic competence, I myself must select the proper term of epistemic
evaluation in terms of which to assess that competence. The term I have settled
on is justified belief, for three basic reasons. First, it appears to be the central
concept of epistemic evaluation that concerns jurists who themselves attend
closely and critically to the process by which scientific expert evidence enters
legal decisionmaking. Second, knowledge, however that concept is cashed out,
seems too demanding for a system that consciously solicits competing expert
scientific testimony (as just noted, judges routinely admit testimony by experts
that is mutually contradictory, with an awareness that two contradictory factual
claims cannot both constitute knowledge). Third, there turns out to be an
interesting and fertile connection between epistemic justification and the
justification required for legal legitimacy.
I shall amplify these three reasons a bit. Over the past several decades,
jurists attentive to the doctrines and institutions of evidence have become

227. See EdmundGettier,Is JustifiedTrueBelief Knowledge?,23 ANALYSIS121 (1963).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1597

increasingly concerned about scientific expert testimony. Something, they

think, is lost along the epistemic chain from scientific research to factfinder
belief. What is it that they think gets lost? It might seem that their concern is
with scientific knowledge possessed by scientists serving as expert witnesses
but lost in the process of transmission through testimony to nonexpert judges
and juries. As a result of this loss, they fear, nonexperts are relying on
unjustified beliefs about scientific information when they render final
decisions. Despite much talk of "knowledge" and "scientific knowledge" in
some of the leading (American) court decisions and litigators' and scholars'
arguments, I maintain that it is not actually knowledge with which these jurists
are concerned, but rather justified belief. (I might put the point this way:
Although many jurists seem to have de dicto concerns about "knowledge" and
"scientific knowledge" and often conduct their analyses in just those terms,
from a philosophical point of view their de re concern is really only with
justified belief.) Daubert is a superb example.
Recall that in Daubert the Supreme Court considered the proper method
by which federal judges are to evaluate proffers of scientific evidence in
deciding whether to admit that evidence for consideration by scientifically
nonexpert factfinders.228The Court framed the issue of the admissibility of
scientific evidence by focusing on the term knowledge, since that term is
central to Federal Rule of Evidence 702, the rule at issue in Daubert.229
Interpreting the "legislatively-enacted Federal Rules of Evidence as [it] would
any statute," the Court proceeded to ascertain what the term knowledge meant
in this rule partly by using the "plain meaning" method of statutory
interpretation (a method that, as in Daubert itself, often amounts to little more
than looking a term up in a dictionary).230The Court's analysis is worth
quoting at length:

The subject of an expert's testimony must be "scientific...

knowledge." The adjective "scientific" implies a grounding in the
methods and procedures of science. Similarly, the word "knowledge"
connotes more than subjective belief or unsupported speculation. The
term "applies to any body of known facts or to any body of ideas
inferred from such facts or accepted as truths on good grounds."
Webster's Third New International Dictionary 1252 (1986). Of course,
it would be unreasonable to conclude that the subject of scientific
testimony must be "known" to a certainty; arguably, there are no
certainties in science. See, e.g., Brief for Nicolaas Bloembergen et al.
as Amici Curiae 9 ("Indeed, scientists do not assert that they know
what is immutably 'true'-they are committed to searching for new,
temporary, theories to explain, as best they can, phenomena"); Brief

228. See supra note 21 and accompanyingtext.

229. See supra note 27.
230. Daubertv. MerrellDow Pharms.,Inc., 509 U.S. 579, 587, 590 (1993).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1598 The Yale Law Journal [Vol. 107: 1535

for American Association for the Advancement of Science et al. as

Amici Curiae 7-8 ("Science is not an encyclopedic body of knowledge
about the universe. Instead, it represents a process for proposing and
refining theoretical explanations about the world that are subject to
further testing and refinement... "). But, in order to qualify as
"scientific knowledge," an inference or assertion must be derived by
the scientific method. Proposed testimony must be supported by
appropriate validation-i.e., "good grounds," based on what is known.
In short, the requirement that an expert's testimony pertain to
"scientific knowledge" establishes a standard of evidentiary
reliability.231

Is the Court really concerned with scientific knowledge? The Court does take
quite seriously the phrase 'scientific knowledge,' but it is clear on reflection
that this cannot quite be the same concept that has concerned traditional
epistemology-at least not those epistemological theories for which truth is a
necessary condition of knowledge. Although the Court and the traditional
epistemologist agree that "the word 'knowledge' connotes more than subjective
belief or unsupported speculation,"232it is obvious that, under the Court's
interpretation, Rule 702 does not presuppose that every expert is testifying to
the truth. Otherwise, the Court would not allow scientific experts to testify to
contrary or contradictory propositions.233 Instead, the Court offers a more
expansive explication of 'knowledge,' one that embraces "any body of known
facts or . .. any body of ideas inferred from such facts or accepted as truths
on good grounds.... Proposed testimony must be supported by appropriate
validation-i.e., 'good grounds,' based on what is known."234(The Court
also concludes that certainty is not a necessary condition of "scientific
knowledge.")235 In a nutshell, the concept of epistemic assessment with
which the Court is concerned when interpreting the term 'knowledge' and the
phrase 'scientific knowledge' is that of a judgment that is supported by good
reasons.

231. Id. at 589-90.

232. Id. at 590.
233. Cf supra Section IV.C (discussing "competition" among experts).
234. Daubert, 509 U.S. at 590 (internal quotation marks omitted). Strictly speaking, one cannot, on
pain of circularity, explicate the concept of knowledge by referring to "any body of known facts," unless
one goes on to give a noncircular explication of what a known fact is. Id. In context, the Court may escape
this problem if what it is really saying is that "when the Rules of Evidence speak of 'knowledge,' they are
actually referring to beliefs supported by good reasons." Even so, the Court speaks a bit too broadly in the
quote in the text above. Surely it is not the case that any "body of ideas inferred from 'known' facts"
satisfies the criteria of even the Court's weaker conception of "knowledge." Id. For there are, of course,
a great many invalid and otherwise unacceptable types of inference possible from acknowledged known
truths. What the Court really has in mind are good inferences (those that are "accepted as truths on good
grounds," "supported by appropriate validation-i.e., 'good grounds,' based on what is known," and the
like). Id.
235. See id. at 590.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1599

We may confirm this interpretive judgment (our interpretive judgment, that

is, regarding the Court's interpretation of the term 'knowledge' in the Federal
Rules of Evidence) by attending to the brief written for the respondents in
Daubert, which clearly had a substantial shaping influence on the Court's
analysis. Like the Court's opinion, that brief also emphasized the distinction
between knowledge in the stronger (philosophical) sense and judgment
supported by good reasons. Indeed, it amplifies this distinction and, even more
to the present point, acknowledges that competing testimony can all be
admitted under the same requirement of "scientific knowledge" in Rule 702:

[E]ven when offering an "opinion," a scientific expert must be

testifying "[]to" "scientific knowledge."
These words naturally mean that an expert must be testifying to
more than merely his or her own view on a scientific issue. Rule 702
does not allow an expert to offer "beliefs" or "hypotheses" or
"theories" or "claims" or "assertions" or "evidence" or "testimony";
instead, it limits expert testimony to "knowledge." Even by itself,
"knowledge" ordinarily requires, at a minimum, appropriate
validation for the proposition-i.e., "good grounds" for the belief
based on what is known. See Webster's Third New International
Dictionary 1252 (1986). And when used as part of the phrase
"scientific, technical, or other specialized knowledge," the term
naturally refers to grounds that are deemed good by the relevant
scientific, technical, or other specialized field-that is, to claims that
are validated, or derived, according to the accepted standards in the
relevant field. See J. Kourany, Scientific Knowledge: Basic Issues in
the Philosophy of Science 112 (1987) (hypotheses "must prove their
mettle" to become "part of scientific knowledge"). Rule 702's
language, naturally read, thus requires validation to the extent
possible-i.e., a foundation, or good reason for acceptance as
valid-based on established standards in the expert's field. 36

236. Brief for Respondentat 14-15, Daubert (No. 92-102) (emphasis added) (second alterationin
original). The brief adds the following footnote:
It is, of course, perfectlyconceivable for each of several competingscientific (or other expert)
claims to be validatedto the extent possible at any given time. The very reason that scientists
can disagree, and that scientific knowledge advances as it does, is that what is known at a
particularmoment does not uniquely predetermineall answers to new questions. Accepted
standards and available evidence thus may not rule out either of two competing, but
well-reasoned,conclusions. By the same token, however,they do rule out some answers.What
validationmeans, therefore-all it can mean, given the ever-evolving body of knowledge-is
good reasonfor acceptance as true, based on what is known at the time.
Id. at 15 n.8. To similar effect are several other passages in the brief. The brief asserts, for example, that
Rule 702 requiresthat "the specific testimonyof each experthave an adequatefoundation,judged by the
accepted standardsof the expert's field," id. at 12 (emphasisadded);that "lilt is the judge's fundamental
duty underthe Rules to screen evidence for admissibilityto ensure that the body of evidence provides a
rationally reliable basis for judgment," id. (emphasis added); that "by restrictingexpert testimony to
'scientific, technical, or other specialized knowledge [that]will assist the trierof fact,' Rule 702 demands
that an expert's testimony be well-groundedin the standardsgenerallyfollowed in his or her field for
validating-establishingthe truthof-assertionsof the type offered,"id. (emphasisadded) (quoting FED.R.
EvID. 702); and that "[t]hecritical term 'knowledge' demandsmore than individualbelief or speculation,

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1600 The Yale Law Journal [Vol. 107: 1535

While this brief's analysis admirably identifies the term of epistemic

assessment that is relevant to rules and institutions of evidence, it also
highlights a deep puzzle about nonexperts' epistemic deference to experts. In
effect, the brief argues that when the putatively scientific evidence proffered
by a party is so weak as not to be "validated, or derived, according to the
accepted standards in the relevant field," then the judge should not even allow
the evidence to be presented to the nonexpert jury. But when evidence
supporting contrary or contradictory propositions is supported by "grounds that
are deemed good by the relevant scientific, technical, or other specialized
field," then the nonexpert judge or jury is to make the decision as to which of
those competing and well-supported claims is to be accepted for purposes of
the legal decision at hand. That is, on this brief's view and apparently on the
view of the Daubert Court itself, when qualified epistemically competent
experts disagree, the decision as to who is correct is to be given by the judge
to the least epistemically competent institutional actor, the nonexpert judge or
jury. Again, we are driven to ask, what is being expected or demanded of the
nonexpert legal reasoner in assessing scientific testimony? The Daubert
opinion and at least some sources on which it relies seem to have it thus:
When the evidence is so weak that no reputable scientist in the field would
endorse it, prevent the nonexpert from hearing it (and from hearing that no
reputable expert would endorse it); but when the best scientific theories and
methods underdetermine the result, let the nonexpert decide who is correct.
How can an epistemically responsible decision emerge from that rule?237
We shall return to this puzzle later. For now, I restate the basic descriptive
observation I have made in this section. What centrally concerns lawyers,
scholars, and judges with regard to the cogency of scientific expert testimony
is not whether the expert has-or can transmit to the nonexpert-knowledge
in the strong philosophical sense, but rather whether the expert has and is in
a position to be able to transmit to the nonexpert a belief that is supported by
good reasons. I think it not inaccurate to go further and say that what concerns
these jurists is not the epistemic concept of knowledge, but rather that of
justified belief. (Obviously a good deal more evidence from the writings of
lawyers, scholars, and judges would be needed to support this claim more
conclusively, but I hazard the judgment that such evidence is readily available.)
I shall go beyond my descriptive claim about what type of epistemic
assessment it is that concerns evidence jurists. Regardless of whether jurists
concerned with the cogency of scientific expert testimony are specifically
focusing on belief supported by good reasons-that is, on justified
belief-rather than on knowledge in the philosophical sense, I now suggest that

but instead refers to inferencesor assertions that are groundedin the standardsreliably used to support
such claims," id. at 14 (emphasisadded).
237. I pose this question, with a differentexample, at the beginning of the Article. See supra text
accompanyingnote 1.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1601

that is the concept with which they should be concerned. Why the weaker
concept only? To anticipate later discussion, while it seems that science has
much of importance to tell the law about matters that are rationally pertinent
to a great many legal decisions, it is also clear that scientific truth is elusive.
That science is not an epistemic monolith or a univocal oracle is trite learning.
In a great many cases to which some particular angle of the scientific point of
view is rationally pertinent, and at every "level" of the scientific point of view
(axiological aims, methods, particular judgments), there is room for skilled,
learned, and reasonable scientists to disagree. Were a legal system to set its
rules of procedure and evidence-the rules guiding "legal epistemology"-so
as to insist on only knowledge (with truth as a necessary condition), the law
would vastly deprive itself of counsel it needs to make legal decisions
sufficiently epistemically legitimate to be legally legitimate. The law is wise
not to have its epistemic reach so far exceed its grasp. Justified belief is all it
does and all it should seek to have transmitted from the scientist-witness to the
nonexpert judge or jury.
But can even that more modest goal be achieved?

B. Testimony as a Source of Justified Belief

I turn now to a more focused investigation of whether the nonexpert can

make the epistemically cogent judgments that legitimate legal decisionmaking
requires-holding off, for the moment, a discussion of the source and nature
of the criteria of legitimacy.238 In accord with the conclusions reached in
Section IV.C, the precise overall philosophical question involved here is this:
Can a nonexpert practical legal reasoner acquire justified beliefs about
scientific propositions and their rational pertinence even in the face of selection
and competition problems? I am attempting to answer that question by
explaining and modeling the reasoning process that underlies epistemic
deference to experts about expert subject matters. What kinds of theoretical
insights are available to help with this inquiry? Insights may be discerned, I
suggest, in two types of philosophical analysis. One is generic epistemological
analysis of testimony as a source of knowledge or justified belief. The other
is more specific analysis by philosophers of the epistemic dimensions of expert
testimony. I consider some of each of these contributions. One terminological
device is worth using in this brief survey. Because these discussions do not
consistently distinguish between knowledge andjustified belief, and because for
my current purposes the distinction is not important, I shall use 'KJB' to refer
indiscriminately to knowledge and justified belief.

238. See infra Part VII.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1602 The Yale Law Journal [Vol. 107: 1535

1. Testimony

I begin with some philosophers' generic analyses of the epistemology of

testimony. Surprisingly little explicit and sustained treatment of testimony is
found in the epistemological literature, given how fecund a source of KJB it
seems to be. In recent years, however, philosophical treatments of testimony
have multiplied, so that the subject will soon likely take its rightful place as
a major domain of epistemology. Much of the contemporary debate focuses on
whether testimony can be an independent source of KJB, on par with memory,
perception, and inference, or whether instead testimony derives whatever
epistemic integrity it has from being reducible to other familiar sources of
KJB.239
Hume offered a brief but influential account in his essay Of Miracles.240
Hume's account is reductionist. He treats the integrity of testimonial KJB as
dependent on empirical confirmation; indeed, his treatment of testimony is a
fairly straightforward application of the thoroughgoing empiricism he
developed in the Treatise241and other works.242 He begins by observing
that "there is no species of reasoning more common, more useful, and even
necessary to human life than that which is derived from the testimony of men
and the reports of eyewitnesses and spectators."243"[O]ur assurance in any
argument of this kind," he maintains,

is derived from no other principle than our observation of the veracity

of human testimony and of the usual conformity of facts to the report
of witnesses. It being a general maxim that no objects have any
discoverable connection together, and that all the inferences which we
can draw from one to another are founded merely on our experience
of their constant and regular conjunction, it is evident that we ought
not to make an exception to this maxim in favor of human testimony
whose connection with any event seems itself as little necessary as
any other.... The reason why we place any credit in witnesses and
historians is not derived from any connection which we perceive a
priori between testimony and reality, but because we are accustomed
to find a conformity between them.244

239. A useful discussion of reductionistand antireductionistviews of testimonyis found in Elizabeth

Fricker,Tellingand Trusting:Reductionismand Anti-Reductionismin the Epistemologyof Testimony,104
MIND 393 (1995) (reviewing COADY, supra note 185).
240. DAVID HUME, Of Miracles, in AN INQUIRY CONCERNING HUMAN UNDERSTANDING 46 (Charles
W. Hendel ed., The Bobbs-Merrill Co. 1955) (1748).
241. DAVID HUME, An Abstract of a Treatise of Human Nature, in AN INQUIRY CONCERNING HUMAN
UNDERSTANDING, supra note 240, at 186.
242. See, e.g., HUME, supra note 240.
243. Id. at 119.
244. Id. at 119-21.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1603

Such an account has a ready explanation of both the reach and the limits
of credulity that one ought, from an epistemic point of view, to extend to
beliefs acquired from testimony: "[W]hen the fact attested is such a one as has
seldom fallen under our observation, here is a contest of two opposite
experiences, of which the one destroys the other as far as its force goes, and
the superior can only operate on the mind by the force which remains."245
Several contemporary philosophers have been attracted to accounts of the
epistemic integrity of testimony that have both elements of the Humean
account: empiricism and reductionism. W.V. Quine and J.S. Ullian, for
example, endorse an account not unlike Hume's, save for the particular
naturalistic and holistic flavor familiar from Quine's general epistemology:

[W]hen we hear an observation sentence that reports something

beyond our own experience, we gain evidence that the speaker has the
stimulation appropriate for its utterance, even though that stimulation
does not reach us. Such, in principle, is the mechanism of testimony
as an extension of our senses. It was the first and greatest human
device for stepping up the observational intake. Telescopes,
microscopes, radar, and radio astronomy are later devices to the same
end.246

Quine and Ullian also provide a naturalistic discussion of the legitimacy of a

hearer's presumption that a "testifying" speaker is telling the truth-telling the
truth both when uttering observation sentences, whose truth the hearer can
check relatively easily247and when uttering nonobservation sentences, which
a hearer often cannot easily check, and thus for which "the danger of mistaken
testimony soars."248Quine and Ullian suggest that, without the presumption
by a hearer of testimony that testifying speakers are telling the truth and
without an extension of that presumption even to nonobservation sentences,
testimony could not be a significant source of KJB. But they also readily
acknowledge that the epistemic integrity of such presumptions is open to
serious doubt, since knavery and fallibility attend both observational and
nonobservational testimony. Even so, they argue that hearers may be justified
in having some presumptive confidence that witnesses are telling the truth, a
justification Quine and Ullian locate in natural features of language and
language learning.249

245. Id. at 121.

246. QUINE & ULLIAN,supra note 148, at 33-34.
247. See id. at 34 ("Observation sentences, taken narrowly, are comparatively foolproof. That is what
makes them the tribunal of science.").
248. Id.
249. See id. at 35. They explain:
Truthfulness is essential, in large part, to the survival of language itself.... Our learning of the
primitive vocabulary of observation sentences consists, after all, in our learning to associate it
with the appropriate sensory stimulations. Small wonder then if those same stimulations dispose
us in the future to affirm the properly associated observation sentences. Lying is an effortful

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1604 The Yale Law Journal [Vol. 107: 1535

Just as there is a natural tendency toward veracity in speakers, argue Quine

and Ullian, there is for related reasons a natural (apparently charity-based,
though they do not say so explicitly) tendency toward credulity in hearers. On
the other hand, credulity is not, and should not rationally be, unlimited.
Rationality reigns in credulity. On Quine and Ullian's view (as on Hume's),
the empirical grounding of the integrity of testimonially acquired beliefs
mandates tight constraints on rational trust in beliefs acquired from testimony:

Veracity is generally admirable, if not always prudent; but credulity,

in more than modest measure, is neither admirable nor prudent.
... The courtroom is worthy of the attention of anyone who is
inclined toward taking too much of what he is told at face value. It
teaches a stern lesson. People disguise the truth in certain situations,
whether out of deviousness, self-deception, ignorance, or fear. They
also, of course, misremember, misjudge, and misreason.250

Neither Quine and Ullian nor Hume explicitly addresses the question of expert
testimony. The difference between general testimony and scientific testimony
can be quite epistemically significant, especially for a Humean account
according to which the integrity of testimony depends on the hearer's ability
independently to confirm or disconfirm what he has been told (at least, in a
sufficiently large sample of cases). Simple, nontechnical (including
observational) reports are easiest for a nonexpert to confirm. A nonexpert
cannot independently and directly check complex theoretical propositions that
do not have simple observational consequences or whose observational
consequences themselves require complex training even to be recognized as
such. Whatever checking the nonexpert can manage must rely on indirect
devices like demeanor, credentials, and reputation.251
Elizabeth Fricker offers a sustained epistemology of testimony that
emphasizes the importance of a hearer's assessment of a witness's sincerity
and competence.252 Although she does not expressly discuss expert
testimony, her emphasis on epistemic competence is illuminating and helpful
here. The main relevant points of her account are these. First, it is important
to distinguish between global and local reductionism in accounts of testimony.
Global reductionism has two forms: (1) the thesis that it is possible to reduce
all testimonial KJB to more familiar, fundamental, and less problematic
epistemic sources and principles, such as perception, memory, and inference;
and (2) the distinct thesis that the epistemologist must make this reduction in

deviation from the conditionedresponse.

Id.
250. Id. at 36-37.
251. I explore these devices shortly.See infra Section V.C.
252. See Elizabeth Fricker, The Epistemology of Testimony, 61 PROC. ARISTOTELIANSOC'Y
SUPPLEMENT57, 73 (1987).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1605

order to account for the epistemic integrity of testimonially acquired KJB.

Hume seems to endorse both of these global reductionist theses;253I surmise,
but am less confident here, that Quine and Ullian do too.254 Local
reductionism also has two forms: (1) that it is possible to reduce some, but not
necessarily all, testimonial KJB to more familiar, fundamental, and less
problematic epistemic sources and principles; and (2) that the epistemologist
must reduce some, but not necessarily all, testimonial KJB in order to account
for its epistemic integrity.
Fricker maintains, rightly, that a great deal of what we are inclined
intuitively to acknowledge as KJB comes from testimony, and that many of
those beliefs cannot be reduced to other sources of KJB. There is powerful
support for this view-much of it marshaled by Tony Coady in his extended
critique and rejection of the reductionist position.255 Against Hume, for
instance, Coady argues that one cannot empirically consult one's personal
experience to check to see whether testimony is reliable because there is too
much testimony and too small a personal observation base. Moreover, much
of what one would do to verify or falsify testimonial evidence is itself suffused
with testimonial information; perhaps even the conceptual framework in which
one describes the world and its objects and institutions has been acquired by
testimony. But, as Fricker stresses, antireductionist arguments like Coady's fail
to distinguish global and local reductionism, and that weakens their analysis
of the independent epistemic merits of testimony. In arguing against Coady's
version of antireductionism, Fricker observes that Coady denies only the
possibility of global reductionism. Fricker allows that we seem to have no
choice but to accept much testimonial information on "simple trust" in our
developmental stages, in which we acquire language and concepts from
parents, teachers, and peers.256But we do have a choice once we have come
of epistemic age. At that later developmental stage, she claims, it is both
feasible and indeed rationally obligatory for mature epistemic actors to adopt
a stringently critical stance toward testimony. Posed in this critical stance, the
hearer of testimony is rationally obliged to focus on the testifier's sincerity and
competence.

253. See HUME, supra note 240, at 119-24.

254. See QUINE & ULLIAN,supra note 148, at 33-34.
255. See COADY, supra note 185, at 79-100.
256. Fricker argues:
Each one of us, in becoming the adult master of our commonsense scheme of things, has been
through a historical process of development during which her attitude towards her teachers and
other informants was one of simple trust.... Bearing in mind the role of teaching by others
whom we trust unquestioningly in our learning of language (which is not separate from learning
about the world), this seems inevitable.
Fricker, supra note 239, at 401. Fricker also suggests that the putative KJB acquired at this stage can be
confirmed in coherentist fashion as the stock of our KJB from all sources, including perception, memory,
and inference, grows. She continues: "[A] belief first acquired through testimony very often gains support
later on both through corroboration by other testimony, and through its coherence with what we learn from
perception, and the empirical theory we base on this." Id. at 410.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1606 The Yale Law Journal [Vol. 107: 1535

In an effort to model the reasoning process of a hearer confronted with

testimony (an effort in much the same spirit as my own in this Article, but
with a different focus), Fricker claims that "common sense semantics" endorses
an inference rule that allows an interpreter to move from the fact that a
testifying witness has spoken a proposition to an endorsement of the
proposition itself.257 That inference rule, which she considers to be an "an
analytic truth about the speech act of assertion," is as follows: (S asserted that
P at t and S was sincere, and was competent at t with respect to P -> P) where
'S is competent with respect to P at t' =df 'at t, S believes p -> p,.258
In Fricker's view, a hearer is rationally obliged to presume both sincerity
and competence, but those presumptions are rathereasy to overcome. Sincerity
"can be assumed unless there are signs of its lack," but "the hearer must
always be scrutinizing the speaker for telltale signs of its absence."259
Similarly, Fricker allows that a presumption of competence "may be assumed
as the default setting."260As she acknowledges, Davidsonian considerations
suggest that some "charitable" presumption by the hearer of the speaker's
competence (and sincerity) may be part of the nature of language; and Coady,
Fricker's main target, leans very heavily on such considerations in his account.
Nevertheless, Fricker's empiricist approach leads her (as it led Hume and
Quine and Ullian) to conclude that a properly rational presumption of
competence is sharply limited:

[D]espite the conceptual constraints on interpretation it remains an

empirical question whether particular speakers, on particular
occasions, are either competent or sincere; one which a self-
consciously rational belief-former will wish to have positive evidence
about, before he believes what he is told. Correspondingly, as to what
structure of justification must support a testimony-belief, it is right to
insist that a hearer must be in a position to know that the speaker is
sincere and competent, and that his being so requires his possession
of some particular evidence pertaining to the case at hand.261

Fricker reveals just how weak she believes this presumption of competence is,
and perforce, how weak are the "conceptual" requirements that a hearer be
charitable toward testimony and testifiers whose competence may be presumed:

[W]ith respect to a subclass of tellings only, viz. those with subject

matters for which commonsense psychological knowledge licenses one
to expect the speaker to be competent about them: such as her name,
where she lives, what she had for breakfast, what is in clear view in

257. Fricker,supra note 252, at 72-73.

258. See Fricker,supra note 252, at 73.
259. Fricker,supra note 239, at 405.
260. Id.
261. Fricker,supra note 252, at 77 (emphasisadded).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1607

front of her, and so forth. Again, the speaker [sic]262 must be

sensitive to indicators of its lack.... A hearer who engages in [this
interpretative task] does not believe what she is told uncritically, and
she has empirical grounds for her trust in her informant.263

To the claim that a stronger presumption of competence is rationally

warranted-say, a presumption of competence in the absence of special
evidence to the contrary-Fricker responds that "[t]he proportion of utterances
which are made by speakers who are either insincere or incompetent is far too
high for this to be an attractive policy."264
Fricker argues for local reductionism. We cannot globally reduce
testimonially acquired KJB because we cannot escape the necessity of relying
on simple trust in our formative years (though even some of the beliefs
acquired at this stage are subject to later coherentist confirmation). But we can,
in our mature years, locally reduce testimonially acquired propositions to other
forms of KJB-perception, inference, and memory-by performing the critical
presumptive interpretive exercise outlined above.
Fricker's account has much to recommend it. Its insights are especially
valuable for their focus on a hearer's need to bring a robustly critical stance
toward both implicit and explicit claims of speaker competence. As I have
argued, such a critical stance is built into the law's epistemology in the form
of rules of evidence and procedure that proceduralize the caution of the
factfinder toward expert (and other) factual claims.265 Fricker's fairly
persuasive account suggests that this kind of proceduralized caution attends
every rational encounter with testimony and not just testimonial encounters in
the highly formalized institutional setting of a courtroom. If she is right, legal
rationality in this domain is continuous with rationality more generally.
Fricker's account is also attractive for its cogent orchestration of three strong
intuitions. First, one cannot, pace Hume, confirm by means of one's own
nontestimonial experience all the testimony one has heard. Second, even many
of the testimonially acquired beliefs that cannot be confirmed by other means,
nevertheless, do seem to produce KJB. Third, there is a limit to rational
credulity, where uncritical trust spills into warrantless gullibility.
Fricker's account does leave important questions unanswered that must be
addressed before her account can provide an adequate account of the rationality
of trust in testimony. She leaves us to wonder, for example, how much of what
we think is testimonially generated KJB is actually KJB, because her

262. I assume this is a misprint for "hearer."

263. Fricker, supra note 239, at 405 (emphasis added). How good the empirical grounds are depends,
of course, on what kinds of empirical reasons there are for the presumption in favor of sincerity and
competence. Fricker does not offer much discussion of those grounds in the pieces I have been discussing,
but has more extended discussion elsewhere.
264. Fricker, supra note 252, at 75.
265. See supra Section I.A.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1608 The Yale Law Journal [Vol. 107: 1535

presumptionabout speaker competence applies only to a relatively trivial

"subclassof tellings"(e.g., the speaker'sname, where she lives, what she had
for breakfast,etc.). Surelythe real power of the intuitionthatKJB arises from
testimonycomes from its abilityto explainthe vast majorityof our beliefs that
are far more complex and that we seem not to be in a good position to
confirm.266 Fricker'saccountleaves largely unexplainedhow and whetherwe
can acquireKJB underthese conditions.
Perhaps it is alertness to this difficulty that leads Frickerto make the
sweeping claim that "the key to the epistemology of testimony is:
disaggregate."267 She elaborates,"Disaggregateboth regardingthe question
of whetherand when we may rightlytrustwithoutevidence, and regardingthe
empirical confirmationof speakers' trustworthiness."268 This disaggregation
principledoes not offer any final answersto the foregoingquestionaboutthe
statusof a greatmany apparentlyunverifiable,testimonially-generated beliefs.
(Can these be real sources of KJB?Is there some way to verify them shortof
becoming experts in the domain of these propositions?)But taken as a
regulatory rule for the project of investigating the epistemic integrity of
testimony,this maxim seems unassailable.Some such principleis guiding my
inquiry here into deference by nonexpert practical reasoners to scientific
experts. The foregoing discussion should suggestjust this: Whatevergeneral
account might be given of the epistemic virtue of testimony as a source of
KJB, it cannotbe assumedthatall testimonialknowledgehas the same virtues.
There are special features of scientific knowledge that make it particularly
difficult to see how they can be sources of KJB for nonexpert practical
reasoners who are constrainedby certain norms of epistemic accuracy and
practicallegitimacy.

2. Expert Testimony

I turn now from general philosophicaltheories of testimonyto accounts

that focus on expert testimony more specifically. Some (for example, those
offered by Hilary Putnam and John Hardwig)269 focus on deference to
expertsin everydaylife; these I referto as "collectivist"accountsof legitimate
epistemic deferenceto experts. Others(for example, the accountsoffered by
Kenny and Coady)270focus more specifically, and more pointedly for my
purposes, on expert testimony in courts. I refer to these as "extra-cameral"

266. John Hardwig vigorously advances this point about the complexity of the testimony that we
intuitively believe produces KJB in hearers.See infra notes 273-285 and accompanyingtext (discussing
Hardwig'saccount).
267. Fricker,supra note 239, at 407.
268. Id.
269. See HILARYPUTNAM,The Meaning of "Meaning,"in 2 MIND, LANGUAGEAND REALITY215
(1975); John Hardwig,EpistemicDependence, 82 J. PHIL.335, 335 (1985).
270. See COADY,supra note 185; Kenny,supra note 64.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1609

accounts of legitimate epistemic deference to experts. Both types of account

defend the strongly held intuition that nonexperts can acquire KJB from
experts-even in the face of the selection and competition tasks discussed
above.27' I shall consider how adequate these defenses are.
"Collectivist" accounts of expert testimony suggest that to understand how
and why epistemic deference to experts can yield KJB,272the theory of KJB
itself must depart significantly from the "individualist" epistemological model
that has long dominated philosophy. That standard model is individualist in
that it treats knowledge solely as the property or capacity of individual
knowing minds, and what is good, reliable, warranted, or true in the way of
belief is analyzed from the individual knower's point of view. Although there
are important exceptions, the dominance of this model cuts across
foundationalist, coherentist, internalist, and externalist theories of knowledge.
By contrast, the collectivist model maintains that creating and transmitting
KJB can be, and is, a collective enterprise, rather than solely an individual one.
Collectivist theorists maintain that KJB is far too complex for an individualist
account to explain fully. Hardwig, a leading recent expositor of this view, puts
the intuition in a compelling way:

I find myself believing all sorts of things for which I do not possess
evidence: that smoking cigarettes causes lung cancer, . . . that mass
media threaten democracy, . . . that my irregular heart beat is
premature ventricular contraction, that students' grades are not
correlated with success in the nonacademic world.... The list of
things I believe, though I have no evidence for the truth of them is,
if not infinite, virtually endless. And I am finite. Though I can readily
imagine what I would have to do to obtain the evidence that would
support any one of my beliefs, I cannot imagine being able to do this
for all of my beliefs. I believe too much; there is too much relevant
evidence (much of it available only after extensive, specialized
training); intellect is too small and life is too short.273

There is indeed reason to believe that an individualist model cannot explain

how we, as nonexperts, acquire KJB about a great many of the things we
firmly believe we do hold as KJB. There is even reason to question whether
the individualist model can adequately explain the KJB putatively possessed
by scientific experts. Consider, for example, that empirical scientists routinely
and increasingly depend on computers in constructing, verifying, and falsifying
their theories. But they are usually not able themselves to verify or even to

271. See supra Section IV.C.

272. I continue to use "KJB"for "knowledgeor justified belief' for the same basic reasonoffered at
the outset of this section.
273. Hardwig, supra note 269, at 335. Since Hardwig does not deny that testimony is a form of
evidence, and he clearly has testimonialevidence for the listed beliefs, we must take his assertionin the
first sentence to mean that he possesses no nontestimonialevidence for them.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1610 The Yale Law Journal [Vol. 107: 1535

comprehend-without a good deal of additional training that they probably do

not have the time to acquire and do not feel the need to obtain-the theoretical
work that guides computer scientists when they build the computers and write
the programs on which the empirical scientists rely. Thus, these scientists may
be said to rely on-i.e., epistemically defer to-the work of computer scientists
without understanding the details of the computer scientists' work. Even so,
one would have to traverse a fair way down the path of skepticism to conclude
that these deferring scientists do not have KJB about, or epistemic competence
within, their own fields just because of their limited understanding of the
computer science that underlies the technology on which they rely.274
What might a nonindividualist account of KJB look like? Hilary Putnam
sketches such a theory in some brief but rich remarks about what he calls the
"division of linguistic labor."275Putnam addresses the question of how
experts and nonexperts in a given society manage to use "expert"terms-terms
that come within the special province of expert theoretical disciplines, such as
the empirical sciences-with the same meaning even though the nonexperts are
not themselves competent to know the meaning of those terms. How, for
example, can nonscientist members of modem English-speaking cultures
routinely use the terms 'water,' 'gold,' and 'beech tree' without knowing the
chemistry, physics, or biology required to understand what such terms really
refer to (so as to be able, say, to distinguish "fool's gold" from gold)?
Putnam proposes a "sociolinguistic hypothesis" to explain this
phenomenon. There is, he claims, a "division of linguistic labor" in which
most members of the society acquire such words as 'gold' and 'water' as part
of their general vocabulary, while only a subclass of those members acquire
the expert methods of recognizing whether a given item is within the scope of
those words or not. That is, even though not himself an expert, an epistemic
pedestrian can use "expert" terms with the same meaning as an expert:

He can rely on a special subclass of speakers. The features that are

generally thought to be present in connection with a general
name-necessary and sufficient conditions for membership in the
extension, ways of recognizing if something is in the extension
('criteria'), etc.-are all present in the linguistic community
considered as a collective body ....

Every linguistic community exemplifies the sort of division of

linguistic labor just described: that is, possesses at least some terms
whose associated 'criteria' are known only to a subset of the speakers
who acquire the terms, and whose use by other speakers depends

274. Hardwig offers a nice example of (what he calls) epistemic dependenceamong physicists. See
id. at 346-47.
275. PUTNAM,supra note 269, at 227.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1611

upon a structured cooperation between them and the speakers in the

relevant subsets.276

Putnam's brief discussion offers at least the start of a promising account of

how a community as a whole can be said to have expert knowledge even when
not every member has the requisite expertise. In a community that includes
both experts and lay people, the expert might be said to exercise epistemic
authority over the meaning of terms whose meaning is discerned only by the
use of specialized expert methods.277
Hardwig takes the collectivist theory much further than does Putnam,
making some remarkable concessions to his individualist challengers along the
way. Hardwig entertains, and comes quite close to endorsing, the idea that a
nonexpert, B, can acquire knowledge of some proposition p (and not the more
inclusive KJB on which I have been focusing) from an expert, A, even under
the following conditions: (1) B has not performed the inquiry capable of
providing evidence for the truth of proposition p; (2) B is not competent, and
could never become competent, to do so; (3) B is not competent to assess the
merits of the reasons that expert A offers for his opinion; and (4) B cannot
understand what p means.
In Hardwig's view, it makes sense even under these conditions to say both
that B's belief is rationally justified and that B knows that p. His principal
argument for the first of these remarkable conclusions (regarding rationally
justified belief) is just that

we must say that B's belief is rationally justified-even if he does not

know or understand what A's reasons [for believing that p] are-if we
do not wish to be forced to conclude that a very large percentage of
beliefs in any complex culture are simply and unavoidably irrational
or nonrational.278

He makes the same basic argument regarding knowledge, pointing to common

cases of scientific practice in which "each researcher is forced to acknowledge
the extent to which his own work rests on . . . work which he has not and
could not ... verify":279"Unless we maintain that most of our scientific
research and scholarship could never, because of the cooperative methodology
of the enterprise, result in knowledge, I submit that we must conclude that the
p is known in cases like this," in which "each researcher is forced to

276. Id. at 228.

277. Putnamuses the term "division of linguistic labor"because of the particularversion of realist
semantics he was urging at the time. One may just as well call it a division of epistemic labor, as he
himself implicitly acknowledges.See id. ("Thisdivision of linguistic laborrests upon and presupposesthe
division of nonlinguistic labor,of course .... [W]ith the increaseof division of labor in the society and
the rise of science, more and more words begin to exhibit this kind of division of labor.").
278. Hardwig,supra note 269, at 339.
279. Id. at 348.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1612 The Yale Law Journal [Vol. 107: 1535

acknowledge the extent to which his own work rests on ... work which he
has not and could not . .. verify."280
There is much intuitive appeal in the collectivist view. It does seem that
nonexperts can, and do, acquire KJB concerning a great many propositions
about complex scientifically articulated subject matters whose warrant can be
ascertained only by the use of specialized tools that they do not possess and
cannot realistically acquire.28' But collectivist accounts like those sketched
by Hardwig and Putnam are just that-sketches-and they too (Hardwig's
most clearly) offer up epistemological pills that are hard to swallow. Basically,
Hardwig offers a collectivist argument that has the structure of a modus
tollens:

(1) Ifjustified epistemic deference to experts is not possible, then

a very large percentage of beliefs in any complex culture is irrational
or nonrational;
(2) It is not the case that a very large percentage of beliefs in any
complex culture is irrational or nonrational;
Therefore,
(3) Justified epistemic deference to experts is possible.

Is this argument compelling? Clearly Hardwig believes (as does Putnam) that
justified epistemic deference is possible. But Hardwig's modus tollens leans
very heavily on a parade of epistemic horribles: "If justified deference is not
possible, then see how irrational our culture would be!" Although such a
parade delights the skeptic and strikes fear in the heart of every epistemically
responsible citizen, modus tollens cannot suffice where explanation is lacking.
Aside from standing on some strong intuitions, for all Hardwig and Putnam
have explained, we have no less justification for running the inference this
way:

(1) Ifjustified epistemic deference to experts is not possible, then

a very large percentage of beliefs in any complex culture is irrational
or nonrational;
(2) Justified epistemic deference to experts is not possible;
Therefore,
(3) A very large percentage of beliefs in any complex culture is
irrational or nonrational.

One philosopher's modus tollens, as the way would have it, is another's modus
ponens. Collectivist arguments like Hardwig's and Putnam's point to a
conclusion that seems compelling: In a complex society, the epistemic whole

280. Id.
281. See supra notes 272-274 and accompanyingtext. This is the intuitionto which I think Fricker
is insufficiently attentive.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1613

is greater than the sum of its parts, and the collective as such can know, or at
least believe justifiably, far more than any member of the collective could
possibly know individually. They point to something we think is true,
something that many would like to believe is true, but without offering any
explanation of how it could be true. Yet it is precisely the account of how
collective knowing is possible that we seek to provide in our philosophical
explanation.282Without more, these collectivist theories do not provide the
necessary analysis.
To be sure, Hardwig's intuition that testimony on such matters does yield
KJB, even under his striking conditions of hearer incompetence, is compelling,
at least prima facie. But is Hardwig entitled, as a matter of epistemological
debate, to his modus tollens presumption, so that the burden of proof is on the
person who would deny that scientific testimony to a starkly incompetent
hearer (as Hardwig concedes many of us to be vis-a-vis the testimony we
receive) can yield KJB in that hearer? In my view, Hardwig is not entitled to
that presumption. Although the intuitions on which Hardwig leans so heavily
are somewhat compelling, we also have compelling experiential reason to
believe that a good deal of testimony, including testimony by experts, is false.
Surely Fricker and Quine and Ullian are on target in this respect. We have too
much evidence of incompetence, dissembling, and epistemically distorting bias
for the burden of argumentative proof to be set as Hardwig seems to claim.
We need not even enter the courtroom to flood ourselves with memorial
evidence of the "stern lesson" that "[p]eople disguise the truth in certain
situations, whether out of deviousness, self-deception, ignorance, or fear. They
also, of course, misremember, misjudge, and misreason."283
In any event, lest my specific focus here on expert scientific testimony
become lost in more general epistemological considerations, I leave my
reaction to collectivist accounts with this contention: Hardwig's modus tollens
can be of no help to a practical legal reasoner who is faced with actually
competing scientific experts.284 Short of making radical revisions to the
logical principles they are willing to accept, nonexperts cannot believe all
scientific experts when those experts testify in contradictory or contrary ways.
Other epistemological accounts of expert scientific testimony, specifically
those that focus on testimony in legal settings, either do not endorse, or at least
do not focus on, collectivist epistemic assumptions. Instead, their central
concern is the manner and setting in which expert testimony is presented to

282. Nozick offers an account of one mode of philosophicalexplanationas the explanationof the
possibility of some state (knowledge,justice, truth)in the face of other apparentlytrue propositionsthat
deny its possibility. See NOZICK,supra note 5, at 8-24.
283. QUINE& ULLIAN,supra note 148, at 37.
284. See supra text accompanyingnote 226 (discussing "actualcompetition").

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1614 The Yale Law Journal [Vol. 107: 1535

nonexperts in a contentious litigative context.285Proponents of what I call

"extra-cameral"accounts believe that although there are obvious problems with
the orderly, epistemically justified reliance on expert testimony in adversarial
legal systems, these are not problems that inhere in the attempt to transfer
expert information from experts to nonexperts. Instead, these are problems that
occur only when expert testimony is presented in an adversarial context.
Accordingly, these accounts propose to remove expert testimony from the
courtroom (the camera) and to have nonexperts defer to such testimony as
neutral testimony by relatively disinterested experts.
Anthony Kenny provides a philosophical theory of this type, one that is
specifically directed to the question of the reliability of expert evidence in
courts. Kenny begins his analysis by specifying four criteria that a putatively
expert scientific discipline must satisfy in order actually to be an expert
scientific discipline. First, it must be consistent.286 Second, it must be
methodical.287Third, it must be cumulative.288Fourth, it must be predictive
and therefore falsifiable, in a special sense.289
Having articulated the conditions that a discipline must satisfy to qualify
as "scientific," Kenny considers the problem of competing expertise in the
courtroom. With regard to novel (putatively) scientific expert methods, he
observes, parties usually battle out in court the question whether the discipline
to which the method belongs is really a science. He argues that because judges
and juries lack the institutional competence to judge whether a new field really
is a science, "the courtroom is not the best place, and the adversary procedure
is not the right method, to decide what is and what is not a science."290Like
many legal commentators,29' Kenny maintains that, within reasonable limits

285. Many of these accounts presuppose some kind of epistemic collectivism, or at least cohere better
with it than with individualism, for the reasons Hardwig highlights. See Hardwig, supra note 269.
286. See Kenny, supra note 64, at 49 (stating that "different experts must not regularly give conflicting
answers to questions which are central to their discipline" though there may be differences about borderline
cases).
287. See id. at 50 ("[T]here will be agreement about the appropriate procedures for gathering
information within the discipline. A procedure carried out by one expert to reach a particular conclusion
is one which must be capable of duplication by any other expert.").
288. See id. ("[T]hough any expert must be able to repeat the results of others he does not have to:
he can build up on foundations that others have built.").
289. See id. ("It need not necessarily predict the future (paleontology does not). But it must predict
the not yet known from the already known ...."). Note that both Kenny and the Daubert opinion articulate
four criteria of science, and they overlap fairly closely on three of them: testing, peer review, and general
acceptance. Kenny has no explicit version of the rate-of-error criterion the Supreme Court endorses.
Compare Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579, 594 (1993), with Kenny, supra note 64, at
49-50. Note also that Kenny offers his criteria as "four criteria which are necessary conditions for a
discipline to be scientific," Kenny, supra note 64, at 49, while the Court assiduously avoids treating any
of its factors as either necessary or sufficient conditions, see Daubert, 509 U.S. at 593.
290. Kenny, supra note 64, at 51-52.
291. Among evidence scholars, John Langbein articulates this position well:
At trial, the battle of experts tends to baffle the trier, especially in jury courts. If the experts do
not cancel each other out, the advantage is likely to be with the expert whose forensic skills are
the more enticing. The system invites abusive cross-examination. Since each expert is party-
selected and party-paid, he is vulnerable to attack on credibility regardless of the merits of his

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1615

of time and money, expert testimony ought to be an effort to discern the truth,
but that the adversary system overwhelms the effort.292This leads to his
principal recommendation:

To remedy the abuses in the giving of expert evidence we should:

... remove from the courts the decision as to whether a nascent

discipline is or is not a science capable of providing expert evidence.
A register should be set up of such disciplines, and those claiming to
have developed a new science should seek admission to the
register.... The essential thing is that the matter should be decided
not by a judge or barrister in haste, but by experts in adjacent
disciplines at leisure.293

Kenny's register would function as a kind of super-credential, the very

existence of which would lend credence to the expert judgments issuing from
those experts who were duly registered. But his and other extra-cameral
proposals294 can neither surmount nor resolve problems of selection and
competition. They perhaps relocate those problems to a different link along the
epistemic chain from expert to nonexpert, but they do not resolve the problems
or explain how the nonexpert practical reasoner can handle selection and
competition in a nonarbitrary manner.295That is, an extra-cameral proposal

testimony.A defense lawyerrecentlybraggedabouthis techniqueof cross-examiningplaintiffs'

expertsin tortcases. Notice thatnothingin his strategyvarieswith the truthfulnessof the expert
testimony he tries to discredit.
Langbein,supra note 73, at 836. Langbeinthen quotes an article on trial strategywhich reads:
"A mode of attackripe with potentialis to pursuea line of questionswhich, by their form and
the jury's studied observationof the witness in response, will tend to cast the expert as a
'professionalwitness.' By proceedingin this way, the cross-examinerwill reap the benefit of
a community attitude, certain to be present among several of the jurors, that bias can be
purchased,almost like a commodity."
Id. (quoting Joseph Ryan, Jr., Making the Plaintiff's Expert Yours,FOR DEF., Nov. 1982, at 12, 13).
Langbeinconcludes:
Thus, the systematicincentivein our procedureto distortexpertiseleads to a systematicdistrust
and devaluationof expertise. Short of forbiddingthe use of experts altogether,we probably
could not have designed a procedurebettersuited to minimize the influenceof expertise.
Id.
292. Kenny notes: "[T]headversarysystem does not fit well with the use of expertsto assist the court.
It leads to dangers that the experts will be more concernedto assist one or other party to win their case
than to assist the court to arriveat the truth."Kenny,supra note 64, at 61.
293. Id. at 61-62. While Kenny'sproposalis limited to judging the questionwhethera new discipline
is a science, the basic problem he identifies affects competing testimony within acknowledgedscientific
fields no less than competing testimony about what is a science. (The formerwas arguablythe situation,
for example, in Daubert.)But thatomission can be remedied,and otherwritersin the "extra-cameral" camp
do so. KennethCulpDavis, for example,has proposedthatCongresscreate"aresearchorganizationoutside
the [Supreme]Court to make studies at the Court'srequest,"so that the Courtcould assign questions of
legislative fact "to a qualifiedstaff for a study or investigation."KennethCulp Davis, Judicial, Legislative,
and AdministrativeLawmaking:A Proposed ResearchServicefor the SupremeCourt, 71 MINN. L. REv.
1, 9, 15 (1986).
294. These include neutralcommissions, such as Kenny proposes;special researchservices; and the
like. See Kenny, supra note 64, at 61-62.
295. I discuss this criterionof arbitrarinessfurtherbelow. See infra Section VII.A.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1616 The Yale Law Journal [Vol. 107: 1535

like Kenny's would succeed only in relocating the competition among experts
from within the courtroom, where the question is either whose testimony to
admit and subsequently whose to credit among those whose testimony has
been admitted, to some antechamber in which the question will be who among
the competing experts is to be appointed to the external commission? What
criteria will be used to select them, and how could this credential advance the
process of reliable deference by a nonexpert to an expert? Without more, we
have no reason to believe that courts (or, for that matter, legislatures) are more
competent to make that judgment reliably than we have reason to believe that
judges and juries can reliably defer to experts inside the courtroom.

C. Point: The Four Possible Routes of Warranted Epistemic Deference by

Nonexperts to Experts

I have just canvassed a few epistemological accounts of testimony in

general and a few accounts of epistemic deference to scientific experts in
particular, but found little help in explaining how the nonexpert practical
reasoner can handle the problems of selection and competition described
above. I now change tack. I shall consider directly the four principal reasoning
mechanisms that nonexperts, unpossessed of epistemic competence, seem
forced to deploy when they defer epistemically to experts. One is the
nonexpert's epistemically substantive judgment about the scientific evidence in
question. A second is what we may call general canons of rational evidentiary
support. A third is the expert's demeanor, either as he appears before the
nonexpert in person or as indicated by such quasi-literary marks as the tone
and authoritative style of written submissions to the court.296The fourth is
the expert's credentials. I shall consider these four methods in the order just
listed.

1. First Route: Substantive Second-Guessing in Practical Epistemic

Deference

Substantive second-guessing of the expert's judgment seems an unlikely

route to rationally cogent epistemic deference. After all, the nonexpert turns
to the expert precisely because the former does not have the substantive
training, and consequent capacity for expert judgment, that the latter has.
Indeed, the more a nonexpert relies on his own substantive assessment of
scientific evidence, the less one can be said to defer epistemically. Though this
method of nonexpert quasi-deference is thus odd to consider at all, I do so for

296. Not infrequently,expert "testimony"is offered in writtenform (for example throughaffidavits).

This is especially likely to be true when the nonexpertjudge, performingthe Daubert "gatekeeping"
function, is assessing the evidence in orderto make the thresholddecision about admissibility.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1617

two reasons. One is that, on the model of epistemic deference developed earlier
in this Article,297 nothing in principle rules out epistemic deference from
epistemic equals or near-equals, so that nothing in principle prevents a suitably
epistemically qualified practical reasoner from second-guessing the expert on
the merits of his testimony. Such second-guessing does narrow the scope of the
deference as deference, but on my account, deference is a matter of degree and
not an all-or-nothing affair. Even so, considering this to be a mechanism of
practical epistemic deference is still odd, because my stipulated concern is
with the practical reasoner who is not epistemically competent in the scientific
subject matter about which the expert testifies, even though the practical
reasoner has decided that the subject matter of that testimony is rationally
pertinent to a case before him and that information from the expert in that
discipline is therefore worth hearing.
But there is a second reason for considering the "second-guessing" as an
option for even the nonexpert practical reasoner in evaluating expert testimony:
At least some prominent legal systems, including the American federal system,
seem to require it or at least come asymptotically close to doing so. That is,
they seem to require that the nonexpert judge select experts, "defer" to experts,
or choose among competing experts, on the basis of an epistemically
substantive judgment about the merits of an expert's proffered testimony.
The Supreme Court's Daubert opinion is a high-profile culprit here.
Simply put, it instructs judges to make their own independent judgment about
the scientific reliability of proffered expert scientific testimony. Recall that
Judge Kozinski was the appellate judge from whose court the Daubert case
went to the Supreme Court, and to whose court the case was remanded. His
words are worth repeating:

[T]hough we are largely untrained in science and certainly no match

for any of the witnesses whose testimony we are reviewing, it is our
responsibility [according to the Supreme Court's Daubert decision] to
determine whether those experts' proposed testimony amounts to
"scientific knowledge," constitutes "good science," and was "derived
by the scientific method."298

To be sure, the Supreme Court's Daubert opinion qualifies the task of the
judge with the declaration that

[t]he inquiry envisioned by Rule 702 is, we emphasize, a flexible one.

Its overarching subject is the scientific validity-and thus the
evidentiary relevance and reliability-of the principles that underlie
a proposed submission. The focus, of course, must be solely on

297. See supra Section III.C.

298. Daubert v. Merrell Dow Pharms., Inc., 43 F.3d 1311, 1316 (9th Cir. 1995); see also supra notes
53-56 and accompanying text.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1618 The Yale Law Journal [Vol. 107: 1535

principles and methodology, not on the conclusions that they

generate.299

The distinction between methodology and results of methodology is an

important one. It is, for example, central to the model of scientific rationality
presented by Laudan, on which I have relied in explicating the notion of
axiological aims and point of view.300Nevertheless, there are at least two
reasons that an inquiry into methodology is no less epistemically demanding
than an inquiry into the application thereof. First, in a significant percentage
of cases, the boundary between methodology and application is indistinct.
Second, even when the methodology is sharply distinguishable from its
application, it is no less likely to be the methodology that presents the
epistemically troublesome barrier to the nonexpert's comprehension than it is
the theory that underwrites and motivates the methodology.
I take it, then, that a solution to the problems of selection and competition
that counsels "make up your own mind, nonexpert" cannot be adequate.
Indeed, because this is so obviously an unsatisfactory solution, it seems likely
that many judges would be led to convert what is on the surface a substantive
inquiry by nonexpert judges-as directed by Daubert-into a form of
deference based on demeanor and credentials. This is predictable from the
armchair, and there is evidence from the practice of courts that this has indeed
been happening in at least some federal courts under the Daubert regime.
Some courts have essentially converted the Daubert test into the old Frye test,
which, in turn, rests on assessing the credibility of persons who have the
"credential" of being members of the "scientific community."'30'

2. Second Route: Using General Canons of Rational Evidentiary Support

Sometimes an expert's testimony is afflicted by a kind of rational

incoherence that a nonexpert can discern even without training in the expert's
field. One of the clearest examples is self-contradiction. Consider People v.
Palmer (Palmer II),302 a criminal prosecution in which the state's expert
medical witness was called to testify about whether the defendant, on trial for
having stabbed his victim to death, was sane at the time of the stabbing. The
medical expert testified that the bizarre behavior the defendant showed around

299. Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579, 594-95 (1993) (emphasis added) (footnote
omitted).
300. See supra notes 143-158 and accompanying text.
301. See, e.g., Zuchowicz v. United States, 870 F. Supp. 15 (D. Conn. 1994); see also Developments
in the Law-Confronting the New Challenges of Scientific Evidence, 108 HARV. L. REv. 1481, 1514 (1995)
(discussing Zuchowicz and its application of Daubert). Although this may be within the spirit of Daubert,
it is inconsistent with its letter, since Daubert expressly refused to make any one criterion of scientific
reliability necessary and expressly allowed that the Frye criterion was a permissible one. See supra note
51 and accompanying text.
302. 543 N.E.2d 1106 (Ill. App. Ct. 1989).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1619

the time of the stabbing was attributable to the defendant's faking his
psychosis, that the defendant was "grossly psychotic" on the day of the
stabbing, that the defendant "was able to distinguish right from wrong and
could conform his conduct to the requirements of the law," and that the
defendant "was not mentally ill."303Reviewing this testimony, an Illinois
appellate court held that the doctor's self-contradictory statements did not
provide any evidence of the defendant's sanity.304
Self-contradiction in testimony, as in other assertions, is a hallmark of
rational incoherence. It is not the only kind, however. Consider, for example,
the events and legal proceedings leading to the Kentucky Supreme Court's
decision in Potter v. Eli Lilly & Co.305 Potter arose from a former
employee's shooting and killing or wounding of several of his coworkers. The
employee had been taking the antidepressant drug Prozac, and the victims or
their estates brought a products liability action against Eli Lilly, the drug's
manufacturer. The precise issue at the trial was whether Prozac was
unreasonably dangerous and defective and whether it caused the employee to
kill or injure the plaintiffs.306At trial there was a considerable amount of
expert testimony about whether Prozac caused the defendant to commit the
shootings. After hearing all the testimony, the jury decided for the
defendant.307
A person who is not an expert in medicine or psychopharmacology, and
who is called upon to assess expert testimony in a case like Potter, could, and
indeed should, heed the distinction between a drug's causing certain behavior
and that drug's not preventing that behavior, and impose a suitably heavy
burden of proof on any expert's argument that does not respect that distinction
or present an explanation in light of it.308Similarly, one need not be an
expert in psychiatry to discern the rational problem in the Palmer II expert's
self-contradictory testimony, and the nonexpert court in that case properly
discounted it severely. The avoidance of self-contradiction and the mindfulness
of the distinction between causation and nonprevention are two examples of
what we may call general canons of rational evidentiary support. One who is

303. Id. at 1107.

304. See id. at 1109. In an earlierrulingreachedlargelybecauseof the weaknessof the state expert's
testimony,the Illinois appellatecourtthat decided Palmer II had held that the defendantwas not guilty by
reason of insanity,thus overturninga jury verdict of guilty but insane. See People v. Palmer(PalmerI),
487 N.E.2d 1154 (Ill. App. Ct. 1985). At issue before the appellatecourt in Palmer IHwas whetherthe
Double JeopardyClause of the U.S. ConstitutionpermittedIllinois to retrythe defendantafter the court's
ruling in Palmer I. The court held that the state expert's self-contradictorytestimonywas so weak thatthe
earlier ruling went to the sufficiency, and not merely to the weight of that testimony, and thus that the
defendantcould not be triedconsistentwith doublejeopardyprinciples.See Palmer 11,543 N.E.2d at 1109.
305. 926 S.W.2d 449 (Ky. 1996).
306. See id. at 451.
307. The trial and appellatecourts ultimately disposed of the case on the basis of a settlementthe
parties reachedbefore the jury renderedits verdict.See id.
308. Jury deliberationsare not reported,and there is no indicationof the extent to which the jury
relied on this distinctionin deciding in favor of the manufacturerEli Lilly.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1620 The Yale Law Journal [Vol. 107: 1535

not expert in a given discipline can sufficiently understand and use canons like
this to assess expert testimony in an expert discipline.309 It is beyond this
Article's scope to enter into detailed discussion of such canons, though such
a discussion would, I think, be of considerable interest and value. Here, it is
enough to identify the existence of such canons, with the two quick examples
presented above, and consider the extent to which such canons might enable
a nonexpert to evaluate expert testimony in a rationally cogent manner.
To answer that question, let us identify a spectrum along which there are
varying degrees of obscurity in failures of rational evidentiary support. The
obscurity of a message is in the ears of its hearer, and the obscurity with
which I am concerned here is the obscurity to a nonexpert of the failure of
rational coherence in an expert's testimony. At one end of the obscurity
spectrum are those failures that are least obscure and are therefore easiest for
a nonexpert to discern in an expert's testimony. At this end, or close to it,
would be the expert's testimony in Palmer II-assuming that the nonexperts
in that case readily understand (as they seem to) that a person cannot be both
"grossly psychotic" and "sane" at the same time. At the other end of the
spectrum are those failures that are most obscure, and are therefore hardest for
a nonexpert to spot in an expert's testimony. Much closer to this end would
be, for example, the testimony of Dr. Bruce Weir, a statistician and population
geneticist who served as a prosecution expert witness in the O.J. Simpson case.
In his original testimony, Weir gave the jury an analysis that failed to account
for certain DNA characteristics in crucial blood samples, characteristics that
were possibly but not definitely present in some of those samples.310 The
result of this error, as Weir later conceded under cross-examination, was that
his analysis of the probabilities that various combinations of randomly selected
individuals would have certain DNA types, was inaccurate and biased (in a
probabilistic sense) against the defendant.311It seems reasonable to regard
this failing of probabilistic statistical analysis as a failing in the rational
coherence of this expert's testimony, but it is a failing that would require a
good deal of sophistication in the expert's field to discern.
With the spectrum identified, we can assess the question: In what
percentage of cases is it likely that nonexperts will be able to evaluate
rationally expert testimony by deploying general canons of rational evidentiary
support? For at least three related reasons, I speculate (and do not claim
greater certainty) that it is only a relatively small percentage. First, it seems
likely that failures of rational coherence in an expert's testimony will most

309. Socrates-at least Plato's Socrates-deployed general canons of rationalevidential supportto

question, challenge, and embarrassmany of the "experts"he encounteredin Athens. See PLATO, The
Apology, in THE LAST DAYS OF SOCRATES 96 (H. Tredennick trans., 1969).
310. For discussion, see D.H. Kaye, The DNA Chronicles: Bad Numbers, Good Lawyering, and a
Better Procedure, available in 1995 WL 564589.
311. See id.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1621

often be closer to the obscure end of the spectrum than to the nonobscure end.
Palmer II seems the unusual case; Weir's testimony in the Simpson case, the
more usual. Second, and relatedly, many expert witnesses are repeat players
in the "game" of giving expert testimony. It would not take long for word to
get out to the trial bar about experts whose testimony has been so unartful as
to appear to a nonexpert, nonobscurely, insufficient to satisfy general canons
of rational evidentiary support. Indeed, one would hope that someone
"qualified as an expert" by a court would know enough about the substance of
her field not to render such testimony. Third, when the failure of the expert
testimony does occur toward the more obscure end of the spectrum, it is left
to a sufficiently tutored opposing counsel, perhaps aided by his own opposing
expert, to point out the failing to the jury; this is precisely what happened with
Weir's testimony. But the more obscure the expert's failure of rational
coherence, the more comprehension of the expert discipline one must have in
order to see that it is a failure. The nonexpert will thus be at a serious
epistemic disadvantage in discerning such failures. This problem will be
exacerbated in the many cases in which, unlike Weir, the expert denies that his
testimony suffered from any failure of rational coherence. In those cases, the
nonexpert will be little or no better off deploying general canons of rational
evidentiary support than he would be in trying to make a substantive
assessment of the expert's testimony. For, in such cases, it will be difficult for
the nonexpert to judge on the merits whether there really is a failure of rational
coherence, because it will require considerable information about the expert's
discipline even to know whether there was such a failure, and this is precisely
the kind of information the nonexpert is, by hypothesis, unlikely to have.312

312. In most of my analysis, I assume for the sake of argumentthatexpertsare sincere and testify in
good faith, but we should not wholly overlook those who are not. Considerthe following speech, by an
engineer who often served as an expert witness to other prospectiveexpert witnesses, showing them the
ropes:
The way I counteractedthe thing, I used anothertechnique.I used the techniqueas [sic] science
as a foreign language. I made a statement to the attorney that absolutely nobody could
understand.Now, what it amounts to, it's going to terminatethe cross-examination,and it's
going to terminateit in a hurry.
I want the jury to understandwhat I say when I feel there are certainconditions.Under
direct examination,the jury understandseverythingthat I say. Under cross-examination,there
are some things I will allow the jury to understandand there are some things which I will not
allow the jury to understand.
If you don't want the jury to understandsomething,then what you do is you answer the
question precisely, you see. If somebody is workingwith a form of inertia,why I use a form
of inertia. I say, "Do you mean the second bolt above the first bolt," you know. Just get into
something which is a very precise way of saying something.
The intervalof minus infinityto plus infinityof X times X, X2,and you know the-no one
is going to be able to do much with that kind of thing.
And he says, "Can you simplify it?" You say, "See, there's too much simplification
already.This is the only way that I can state it to you so there will be no misunderstanding."
Sanchez v. Black Bros. Co., 423 N.E.2d 1309, 1320 (III.App. Ct. 1981) (emphasisomitted) (determining
that a trial court's refusal, during cross-examinationof a manufacturer'sexpert witness, to permit
questioningabout this speech was reversibleerror).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1622 The Yale Law Journal [Vol. 107: 1535

3. Third Route: Evaluating Demeanor in Practical Epistemic Deference

As I discussed in Section V.B, several philosophers have remarked on the

importance of a speaker's demeanor as an epistemic tool by which a hearer
assesses and evaluates that speaker's testimony. Fricker makes the rebuttable
presumption of sincerity on the part of a witness a centerpiece of her
account.3"3Were Quine and Ullian to offer a more thoroughgoing account
of testimony consistent with Quine's general epistemology (an ambition they
seem to have in the joint work314), they would surely argue that it is both
possible and necessary for a language-learner to distinguish sincere assertions
of belief by an expert from insincere ones. Hume, too, was sensitive to the
dynamics of demeanor in his treatment of testimonially acquired KJB.315
There is indeed a proper epistemic role for judgments of demeanor in the
assessment of testimony, both expert and nonexpert. One might even go so far
as to conclude that a hearer's capacity to assess accurately the demeanor of a
witness is a necessary condition of the acquisition of KJB from that witness.
The hearer should be able to discern whether the speaker is dissembling, and,
when she is dissembling, the hearer will usually decline to endorse the
testimony reported. Even when the hearer discerns that the speaker is
dissembling, however, he need not reject her testimony, for it is possible that
the prevaricating witness may bungle the effort to lie in a way that still allows
the testimony to provide useful evidence.
But assessment of demeanor is unlikely to be accurate enough in general
to provide a basis for an explanation of how a nonexpert can acquire KJB from
an expert. Demeanor is an especially untrustworthy guide where there is what
we might call a lucrative "market"for demeanor itself-demeanor has "traded"
at high prices since the days of the sophists and finds exceptionally robust
business in adversarial legal systems. When judges and juries use demeanor as
a test for the credibility of expert evidence, they face this severe difficulty:
Epistemic warrantand persuasiveness diverge, especially when the "persuadee"
has too limited an epistemic capacity to be able to assess competently the
epistemic warrant of testimony independently of the criteria that make an
expert seem persuasive.
Demeanor is perhaps the chief of these criteria, as Aristotle explained.316
He maintained that there are three principal means of persuasion-appeal to
reason (logos), appeal to emotion (pathos), and appeal to the character of the
speaker (ethos)-and that, of these three, ethos was the most effective, more

313. See Fricker,supra note 239.

314. See QUINE & ULLIAN, supra note 148, at 33-41.
315. See HUME,supra note 240, at 120 ("Weentertaina suspicionconcerningany matterof fact when
the witnesses ... deliver their testimony with hesitation or, on the contrary, with too violent
asseverations.").
316. See ARISTOTLE,RHETORIC75 (A. Jebb trans.,CambridgeUniv. Press 1925).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1623

effective than appeal to reason (not surprising, perhaps) or appeal to emotion

(surprising, but the point does survive careful reflection). The demeanor of an
expert witness is precisely what Aristotle referred to as the ethos of that
"speaker."But as theorists of KJB, we have no reason to believe that an expert
witness's persuasive demeanor has any particular connection to the epistemic
warrant for what the witness asserts. This is the basic epistemic obstacle to the
use of demeanor as a route to legitimate practical epistemic deference in legal
systems (like the American) in which there is a market for demeanor. Judges,
lawyers, and commentators are thoroughly aware that lawyers choose expert
witnesses at least as much because they will appear to a jury to be competent
as because (in the lawyer's judgment)3"7the experts actually are competent.
A brief look at some of the actors in this market is instructive.318
A 1967 survey of judges, lawyers, and doctors in the Los Angeles area
found that "[o]ver three-quartersof the attorneys responding . . . indicated that
some factor other than medical expertise-usually an impressive 'courtroom
manner'-often determines the choice of an expert witness."319An article in
a litigator's trade magazine describes the selection of experts as follows:

Usually, I like my expert to be around 50 years old, have some

grey in his hair, wear a tweedy jacket and smoke a pipe ....

You must recognize the jurors have prejudices, and you must try
to anticipate those prejudices ....
Some people may be geniuses, but because they lack training in
speech and theater, they have great difficulty conveying their message
to a jury.320

Another trade magazine instructs that an expert witness must "exude

confidence, create empathy, and seem and be completely sincere and
convincing."321 A law review article similarly notes that "the selection
process involves more than securing an expert who will render a favorable
opinion. The credibility and persuasiveness of an expert are equally important
concerns."322 Another author adds that "[t]he best expert is one... who

317. Notice that when the judgmentof the lawyer is also nonexpert,as it often is, the lawyer may not
even be in a good position to decide on substantivegroundswhich expertsare in fact competent.That is,
the lawyer's relativelack of epistemiccompetencecompoundsthe problemof possible divergencebetween
warrantedassertionand persuasiveness.
318. I draw on materialcollected in Samuel R. Gross, Expert Evidence, 1991 Wis. L. REv. 1113,
1126-36. Similar materialsare gatheredin Langbein,supra note 73, at 835-41.
319. Note, The Doctor in Court: Impartial Medical Testimony, 40 S. CAL. L. REv. 728, 728-29 (1967).
It is not clear whether the authors of the survey were attentive to lawyers' relative inability to judge the
epistemic merits of experts' testimony.
320. Hyman Hillenbrand, The Effective Use of Expert Witnesses, BRIEF, Fall 1987, at 48, 49 (emphasis
added).
321. Stephen E. Nagin, Economic Experts in Antitrust Cases, LITIGATION, Winter 1982, at 36, 37.
322. Thomas V. Harris, A Practitioner's Guide to the Management and Use of Expert Witnesses in
Washington Civil Litigation, 3 U. PUGETSOUND L. REv. 159, 161 (1979).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1624 The Yale Law Journal [Vol. 107: 1535

has . . . qualities that give a certain 'glow' to an otherwise acceptable position.

In fact, the best expert testimony in the world may be utterly useless unless it
is presented by someone whose other attributes can add a ring of truth to
it."323 Lawyers are of course being prudent when they choose experts who
will favor their side and who will do so with convincing demeanor. But
because, ex hypothesi, nonexpert factfinders do not have epistemic competence,
they are not in an epistemic position to be convinced by substantive
arguments-that is, they cannot be convinced by what Aristotle called the
reason (logos) behind an expert judgment because they cannot understand those
substantive arguments.324 In an adversarial setting, the criterion of being
convincing to the nonexpert subordinates the criterion of being competent to
produce accurate expert scientific judgments.325My point here is not that the
market for demeanor is biased in selection toward false or ignorant experts.
Rather, I am suggesting that given what the market is selecting for (something
that will convince jurors who are not competent to judge substantive scientific
argument), there is no reason to believe that epistemic warrant has any
particular connection with what chosen experts will say.

4. Fourth Route: Evaluating Credentials

Credentials (along with reputation, which I treat as a species of credentials)

is probably the most important device relied upon by philosophers and jurists
to explain how practical epistemic deference can yield KJB. Collectivist and
extra-cameral accounts of epistemic deference share the assumption that
nonexperts can use credentials to acquire KJB from experts. This same basic
assumption also motivated the Frye rule, a rule that many evidence scholars
still support. When coupled with a fairly accurate capacity to use an expert's
demeanor as a guide to his sincerity, the nonexpert's judgment of credentials
seems the most promising means by which to explain how a nonexpert can
acquire KJB from an expert scientist.
I will now investigate this mechanism to see whether it is likely to produce
KJB in a nonexpert. I assume for the purpose of this investigation that the
epistemic device of credentials is coupled to that of demeanor, and that the
nonexpert who uses demeanor "evidence" is a fairly accurate judge of when
the expert is being sincere. The simple reason for these related assumptions is
that if indiscernible prevarication were a habit among experts, credentialism
alone would be incapable of generating KJB. By assuming that the credentialed

323. Albert Momjian, Preserving Your Witness's Stellar Testimony: How To Qualify Your Expert to
the Court, FAM. ADVOC., Summer 1983, at 8, 8.
324. Cf. supra Section IV.A (discussing understanding and epistemic competence).
325. Samuel Gross makes a related but slightly different point: "The confident expert witness is less
likely to have been chosen because she is right, than to have been chosen because she is confident whether
or not she is right." Gross, supra note 318, at 1134.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1625

expert is sincere and is accurately perceived as sincere by the nonexpert, I am

able to consider credentialism as a potential route to KJB in the setting in
which credentials are likely to be most helpful.
The use of credentials (assuming sincerity) as a route to practical epistemic
deference has both strengths and weaknesses. I begin by mentioning some of
the weaknesses and then setting these against the strengths. There are at least
three closely related reasons that use of credentials seems unlikely to provide
the nonexpert practical reasoner with KJB: regress, question begging, and
underdetermination. These reasons are so closely related that traversing the
path along any one of them soon leads to one or both of the others.

a. The Regress Problem

Regress can involve one or more of the selection problems, or one or more
of the competition problems identified in Section IV.C. Recall that the
selection problems are: (1) determining which of the intellectual enterprises
that might yield expert testimony is a science; (2) determining who is a
scientist capable of using her science in a manner that satisfies the standard of
epistemic appraisal and the attendant level of confidence that the practical
reasoner has established; (3) determining which of the intellectual enterprises
that might yield expert testimony is a science that is rationally pertinent to the
case; and (4) in cases in which there is significant doubt occasioned by task
(3), determining who is capable of answering (3) in a way that can identify an
expert scientific discipline capable of satisfying the chosen standard of
epistemic appraisal and the attendant level of confidence. For the purposes of
this discussion, it is not important to determine exactly which of these
selection problems the nonexpert faces, for the regress problem can arise
regarding each one (and several, though perhaps not all, combinations of
them).
The basic problem of competition is how a nonexpert can rationally decide
which of the competing experts (whose competition is intra-disciplinary or
extra-disciplinary, actual or implied) to believe when the nonexpert is not
competent to assess the substantive merits of the experts' competing
arguments. When experts testify to contrary or contradictory propositions, the
nonexpert must decide whom to believe on the scientific issue. But, ex
hypothesi, the nonexpert does not have sufficient competence in the expert
discipline to be able to make the choice on substantive grounds, so on what
rational basis can the nonexpert make that choice?
A solution to these related problems commonly offered by both jurists and
philosophers is to maintain that nonexperts can and do acquire KJB from
experts by relying on credentials. Kenny's solution to the problem of expert

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1626 The Yale Law Journal [Vol. 107: 1535

testimony, for example, is a kind of super-credentialism.326Although Putnam

and Hardwig are less explicit about it, their nonexpert epistemic communalists
would presumably consult credentials to identify the "scientist" whose
referential terms are to be deferred to (Putnam) or who is the expert to be
accorded deference in the collective (Hardwig).327This same basic approach
underlies the Frye rule that Daubert displaced,328 and even after Daubert
many federal and state courts have gravitated back to it.329Typical of post-
Daubert endorsement of Frye's "credentialist" solution to the problem of
expert evidence is this assertion by an evidence scholar:

Science is the only source of its own reliability. Anything less than
complete deference to the weight of credible scientific opinion
concerning the reliability of scientific evidence means going outside
science-to the judge or jury . . . to resolve a scientific dispute. The
resulting judgment cannot be scientific and therefore we cannot
honestly speak of the evidence as having "scientific" reliability....

. . .[T]he "real" issue is whether good scientists consider the

evidence reliable at this time.330

These commentators argue that the more modest task that the credentialist
solution sets for judges331is much better suited to their limited capacities to
understand complex scientific evidence.332 Whereas Daubert requires courts
to judge whether some given evidence is scientifically reliable, what I call the
"Frye solution" is to have judges ask whether scientists think that the evidence
is reliable.
But which credentials indicate membership in the scientific community?
An Ed.D.? A Th.D.? A Ph.D. from a correspondence school? A degree in
"creation science"? Analogously, would a Ph.D. in philosophy from such an
unlikely place as the University of Pittsburgh333be worthy of respect were
one looking for philosophical expert testimony?334 Clearly, in an age in

326. See Kenny, supra note 64, at 61-62.

327. See supra notes 273-286 and accompanyingtext.
328. See supra Section I.B.
329. Daubert made the Frye test one of its four factors of scientific reliability.Some federal courts
have resuscitatedthe Frye test while ostensibly applying the Daubert rule, and many state courts have
expressly rejectedthe Daubertruleand expressedcontinuingallegianceto the Frye test when applyingstate
rules of evidence. See Developmentsin the Law-Confronting the New Challengesof ScientificEvidence,
supra note 301, at 1514 n.40 (citing cases).
330. Milich, supra note 68, at 923-24 (emphasisadded).
331. I say "judges"because at the stage of litigationto which Frye is relevant,the question is one of
admissibility.Obviously,at a laterlitigativestage, factfindingjuries can also rely on credentialsin assessing
the weight to be given to evidence that a judge has seen fit to admit.
332. See, e.g., Milich, supra note 68, at 918-20.
333. Most jurors presumablywould give less weight to a degree from Pittsburghthan from, say,
Harvard,since they would be ignorantof Pittsburgh'slofty reputationin philosophy.
334. Cf. supra note 60 (discussing a case in which philosophers'testimonywas taken).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1627

which formal credentials have significant market value in many different kinds
of markets, the nonexpert needs guidance through the thicket of would-be
experts, wannabe experts, and magic elixir mongers. Here, a new solution
presents itself (actually, once again, it is the old solution relocated): Have the
nonexpert judge or jury consult some kind of "meta-expert" (someone who is
an expert about expertise in a given area) for a list or specification of the
credentials that a nonexpert could reliably use to pick out competent experts
in the scientific field. But then how can the nonexpert rationally identify the
proper "meta-experts"? On what basis is the nonexpert to identify those meta-
experts who have KJB about the proper credentials? Must not the nonexpert
rely on credentials (including reputation) to identify the appropriate meta-
experts as well?
Thus, it seems that the "Frye solution"-ask those with the credentials of
science whether an expert has the credentials of science-threatens to slide into
an epistemically unworkable regress. Using credentials only pushes the inquiry
back a step without resolving the basic problem. If nonexpert judges and juries
are not competent to judge the content of expert information, how are they
going to be competent to judge credentials of those who would give expert
information? If the answer is to ask credentialed "meta-experts" what the
proper credentials are, the regress has begun. Yet the "Frye solution," so
common in one form or another in analyses by jurists and philosophers, cannot
let the regress slide on infinitely if these analysts truly intend to explain how
the nonexpert's use of credentials can serve as a possible or actual means of
acquiring KJB from an expert. Nonexperts certainly cannot indefinitely
continue to ask expert after expert about proper credentials. Somehow, the
regress must be stopped.

b. The Question-Begging Problem

Can the nonexpert stop the regress without vicious question begging? In
many-I do not say all-cases, credentials will be of little help, for they will
either reproduce the problem (leading to regress) or cause the nonexpert to
settle on one competing expert without having a good reason (thereby begging
the question). We have seen how the credentializing solution (like the Frye
solution) can lead to regress. Such a solution can also lead to question begging,
as the following example indicates.
McLean v. Arkansas Board of Education335was one of the early federal
creation science cases dealing with the constitutional merits of a statute that
mandated equal classroom time for evolution and "scientific creationism."
Relying on expert testimony, the district judge concluded that "scientific
creationism" could not pass muster under the Establishment Clause because it

335. 529 F. Supp. 1255 (E.D. Ark. 1982), aff'd, 723 F.2d 45 (8th Cir. 1983).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1628 The Yale Law Journal [Vol. 107: 1535

was not science but a religious doctrine that public schools could not
promulgate. But assuming that the judge was not an expert on science (or, for
that matter, philosophy or religion), whom should he have asked about the
identity of the appropriate scientific expert on whether creation science is real
science-a creation scientist or a Darwinian? Either answer begs the
question.336 Moreover, to make matters at the very least somewhat more
difficult for the nonexpert, "creation scientists" are quite careful to proclaim
their own credentials as scientists. Indeed, the very use of the terms 'creation
science' and 'Christian science' are efforts at establishing credentials. Philip
Kitcher criticizes creation scientists for just this kind of "credential
mongering":

Creationist claims about credentials look better when presented in soft

focus.... On closer inspection, the "21 scientists who believe in
Creation" [listed in a creationist publication] hardly constitute a
distinguished panel of experts on the origins of life: three hold
doctorates in education; two are theologians; five are engineers; there
is one physicist, one chemist, a hydrologist . .. one entomologist, one
psycholinguist, and someone who holds a doctorate in Food Science
Technology; finally, there are two biochemists . . . an ecologist, a
physiologist, and a geophysicist. While the last five may have some
expertise in related areas, the credentials of the others are utterly
irrelevant to many of the questions Creationists address. The
"authority" of these men should not convince us that there is a
scientifically reputable alternative to a major biological theory. The
word of just any "scientist" is not enough. I am prepared to bet that
Creationists, like the rest of us, take care to consult the appropriate
experts. I doubt that they take their sick children to the vet.337

One ought to concede to Kitcher that it would be far too quick a skepticism
to conclude from the possible fact of disagreement that no one can ever know
the truth of the matter or have decisively good reasons that overcome other
reasons, and I do not draw that conclusion here. That is, nothing in my
argument denies that there are good, or even decisive, arguments on such
questions as whether creation science is "real" science.
But my current concern is not whether there are compelling
arguments-expert arguments such as might be made by a philosopher of

336. Of course, there are versions of this dilemma that pit traditionalscience (the science of
"established"universitiesand learnedsocieties) against "nontraditional" science more directly.There are
cases in the Americancourts,for example,dealingwith the desireof ChristianScientistparentsnot to allow
their gravely ill childrento receive conventionalmedical care. See, e.g., Newmarkv. Williams, 588 A.2d
1108 (Del. 1990). These cases turnin parton whetherthe child is actuallyreceivingadequatemedicalcare.
Whom should the judge ask about whether ChristianScience healing methods have due regardfor the
traditionalmedical facts, or indeed whetherthe traditionalmedical facts are the only relevantfacts to be
known-the ChristianScientist or a memberof the traditionalmedical establishment?
337. PHILIPKITCHER,ABUSINGSCIENCE:THE CASE AGAINSTCREATIONISM 179 (1982).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1629

science-for the view that creationism is only ersatz science. Rather, my

concern is with a nonexpert's epistemic capacity to use credentials as a
rationally warranted means of resolving that debate. Kitcher himself helps to
make my point on this separate question. After offering the remarks quoted
above, Kitcher argues that "the crucial issue is not whether some people who
possess doctoral degrees say that there is a case for Creationism, but whether
they are right in saying so. To settle that issue, it is wiser to look at the
evidence itself."338Very well, but now we have come full circle, for looking
at the nontestimonial evidence to decide who, among competing "credentialed"
experts is right, is precisely what the nonexpert is incapable of doing. From the
point of view of the nonexpert, relying on Kitcher or Stephen Jay Gould as
experts is no less question begging than relying on "Dr. Henry Morris" who
proclaims himself to be "recognized as one of America's greatest authorities
on scientific creationism," who is "[a]rmed with three earned degrees
(including a Ph.D.) in the sciences," and who "served as department head or
professor at four famous institutions, Louisiana University, the University of
Minnesota, Rice University, and Virginia Polytechnic Institute."339My only
contention is that nonexperts are not in a position to judge who is in a position
to judge such matters without begging the question. Whether such question
begging is vicious, harmless, or even virtuous, awaits discussion of the role of
nonarbitrariness as a practical aim governing the legal reasoner's epistemic
deference to experts.340I should also hasten to add that, though my example
may be an unusual one, it is actually not unrepresentative of the kind of
epistemic predicament in which the nonexpert practical reasoner typically finds
himself.341

338. Id.
339. HENRYM. MORRIS,THE REMARKABLE BIRTHOF PLANETEARTHat back cover (1972), quoted
in KITCHER,supra note 337, at 178-79.
340. See infra Part VII.
341. One example (and there are many like it) is found in the following case, which addressed the
same issue as in Daubert, namely whether Bendectin could cause birth defects:
Have the plaintiffs established by a preponderance of the evidence that ingestion of Bendectin
at therapeutic doses during the period of fetal organogenesis is a proximate cause [i.e. does it
in a natural and continuous sequence produce injuries that would not have otherwise occurred]
of human birth defects? . .. The jury unanimously answered no. Judge Rubin denied a post-trial
motion for j.n.o.v. by the plaintiffs because "[b]oth sides presented testimony of eminently
qualified and highly credible experts who differed in regard to the safety of Bendectin." The
great weight of scientific opinion, as is evidenced by the FDA committee results, sides with the
view that Bendectin use does not increase the risk of having a child with birth defects. Sailing
against the prevailing scientific breeze is the DeLucas' expert Dr. Alan Done, formerly a
Professor of Pharmacology and Pediatrics at Wayne State University School of Medicine, who
continues to hold fast to his position that Bendectin is a teratogen. In spite of his impressive
curriculum vitae, Dr. Done's opinion on this subject has been rejected as inadmissible by
several courts.
DeLuca v. Merrell Dow Pharms., Inc., 911 F.2d 941, 945-46 (3d Cir. 1990) (citations omitted). In a
footnote, the court added:
Dr. Done served as a Special Assistant to the Director for Pediatric Pharmacology of the FDA's
Bureau of Drugs from 1971 to 1975. In this role, Done aided in the provision "of FDA input
on research involving children and fetuses, and development of guidelines for pre-clinical safety

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1630 The Yale Law Journal [Vol. 107: 1535

c. The Underdetermination Problem

It is really underdetermination that gives rise to both the problem of

regress and the problem of question begging rehearsed above.
Underdetermination creates a difficulty for credentialist approaches to
explaining the legitimacy of epistemic deference to experts: When the
credentials of the experts are, to the eyes of the nonexpert, evenly matched for
all the nonexpert justifiably believes-that is, when they underdetermine the
credibility of the competing witnesses-it is very difficult to see how
credentials could provide an epistemically legitimate method the nonexpert can
use to resolve selection and competition problems.

D. Counterpoint: The Anti-Skeptical Response and the Dialectical Impasse

The foregoing skeptical considerations,342however, may seem too quick,

too cheap, too thin. It is unduly skeptical to deny that nonexperts can ever use
credentials as a tool to acquire KJB on the basis of epistemic deference.
Surely, even nonexperts can judge credentials well enough to handle what
might be called the crank factor,343 also familiarly recognized as the Flat

evaluations of drugs for use in children and in pregnancy..." He also participatedin

publishinga papercalled "GeneralGuidelinesfor the Evaluationof Drugs to be Approvedfor
Use duringPregnancyand for the Treatmentof Infantsand Children,"in conjunctionwith the
American Academy of Pediatricsin 1974. Dr. Done's opinion that Bendectin is a teratogen
largely rests on inferenceshe drawsfrom epidemiologicaldata, most of which he contendsare
the same that was utilized by the experts,includingthe FDA committee,to whom MerrellDow
cites to bolster its contention that Bendectin does not cause birth defects. The principal
difference is that Dr. Done analyzes that data using an approach,advocated by Professor
KennethRothmanof the Universityof MassachusettsMedical School, that places diminished
weight on so-called "significancetesting."
Id. at 946 n.7 (citationsomitted);cf. K.J. Rothman,A Show of Confidence,300 NEw ENG. J. MED.1362
(1978) (arguingthat significancetesting is misleading).
342. Concernsaboutregress,questionbegging,andunderdetermination arepresentin ancientskeptical
(specifically, Pyrrhorian)challenges to the claim that we can know anything. See JULIUS ANNAS &
JONATHANBARNES, THE MODESOF SKEPTICISM19-30 (1985) (discussing "ten modes" of Aenisidemus).
343. Judge Posner describedthe crankfactor with typically unvarnishedasseveration:
[The witness' testimonywas] either [that]of a crankor, what is more likely, of a man who is
makinga careerout of testifyingfor plaintiffsin automobileaccidentcases in which a door may
have opened;at the time of trialhe was involved in 10 such cases. His testimonyillustratesthe
age-old problem of expert witnesses who are "often the mere paid advocates or partisansof
those who employ and pay them, as much so as the attorneyswho conduct the suit. There is
hardlyanything,not palpablyabsurdon its face, that cannotnow be proved by some so-called
'experts."'
Chaulk v. Volkswagenof Am., Inc., 808 F.2d 639, 644 (7th Cir. 1986) (Posner,J., dissenting) (quoting
Keegan v. Minneapolis& St. L.R.R., 78 N.W. 965, 966 (Minn. 1899)). One can be forgivenfor wondering
to what extent Posner's view is driven not so much by a concern for the expert witness's epistemic
competence as by a strong disagreementwith the expert over the normativeattractivenessof the expert's
views:
It is not the law, in Wisconsinor anywhereelse, thatthe standardof care is set by the designers
of $60,000 automobiles,so that the omission of any safety device found in such automobiles
is negligent.... The buyer of a Mercedes 560 may be willing to pay extra for minuscule,
perhapswholly theoretical,improvementsin safety, but such a buyer's willingness to buy the
ultimate refinementin safety technology does not define the standardof care for the whole

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1631

Earth Society phenomenon. Presumably the admission of crank science is, after
all, what Daubert seeks to reduce by having the trial court judge play a
"gatekeeping" role, using his assessment of scientific reliability as a criterion
of admissibility. The skeptic, however, will insist in response: But how often
can a nonexpert justifiably dismiss an expert's credentials as cranky? Even if
the nonexpert's use of credentials could help weed out cranks by weeding out
egregiously uncredentialed putative experts, in a great many cases the
credentials of experts will remain-to the eyes of the nonexpert-evenly
matched.
Let us suppose with the anti-skeptic that there is an epistemically valuable
role for credentials in a nonexpert's assessment of expert testimony. What
might that role be? Coady provides a modest and reasonable answer:

We have certifying bodies and institutions and their various

certificates and, typically, the courts require that the [expert] witness
be shown to have some relevant certification from such bodies. Doubt
can arise, of course, about the credentials of supposedly expert
institutions ... but usually the courts do not doubt such credentials.
Were they to require for every such certifying body some proof of its
credentials, it is hard to see what could be forthcoming, other than
more of the same.3"

Universities, colleges, and scholarly associations are the leading examples of

"certifying institutions" that play the credentialing role. Certainly in our
everyday lives and work we make many judgments on the basis of credentials
from these institutions. And certainly many feel that this manner of
credentialism works fairly well in those spheres. Our teeth get their cavities
replaced; our computers get their hard drives fixed (and built in the first place)
and run programs that perform wonders-wonders verifiable to all but
Cartesian skeptics; our microwave ovens verifiably cook food in very few
seconds; our cars and planes verifiably carry us at high speeds or altitudes; our
home entertainment systems verifiably deliver larger and larger sights and
sounds from smaller and smaller systems of audio and visual information. We
are aware, indeed, too often reminded, that behind each technological miracle
stands a network of credentialed inventors, technicians, and theorists. All of
this is powerful evidence that something is not rotten in the world of
credentialed epistemic authorities.

industry.
Id. at 644-45. This is perhapsanotherinstance in which the borderlinedemarcatingthe expert's zone of
competence is blurryand in need of policing. It is also relevantto the question of underdetermination
to
note thatthe Chaulkmajoritycharacterizedthe expertas "anengineerwith expertisein automobilesafety."
Id. at 642.
344. COADY, supra note 185, at 282.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1632 The Yale Law Journal [Vol. 107: 1535

Still, questions remain. Coady frankly acknowledges one problem with

such certifying institutions-their "inbuilt tendencies towards intellectual
conservatism and towards the monopolization of social control over knowledge
conceived of as a kind of commodity."345That is certainly an important
problem, assuming that commodified monopolization of this sort has an
epistemically distorting effect, as it surely can. (It is in effect a problem of
competition, either intra- or extra-disciplinary, actual or implied.) How is the
nonexpert to become aware of, or resolve, this particular problem of
competition? Coady offers little in the way of a persuasive answer. He
suggests that "the courts should give initial credence to the verdicts of such
bodies as universities on the issue of bogus science, though they should be
But
prepared to hear argumentabout the matter if it can be produced."346
what kind of argument should they be prepared to hear, and how would they
assess it if they heard it? What is Coady supposing the nonexpert can do in
making such an appraisal-that she can do a better job of adjudicating between
competing experts (including would-be monopolists and those seeking to raise
the barriers to epistemic entry) than the experts themselves can do?347As
long as judges and juries lack epistemic competence, what reason do we have
to believe that they will be able to make a justified judgment about the
arguments they are "prepared to hear" that contravene the judgments of the
certified experts?
Before moving on, let me sum up the problems that remain for
credentialism as a proposal for solving the related problems of selection,
competition, and underdetermination. Coady's solution is to have nonexperts
give prima facie weight to credentials-principally, memberships in reputable
universities and learned societies. This solution could work only if the weight
of credentials is clearly on one side; but often the nonexpert justifiably believes
that credentials are not so clearly weighted. The underdetermination difficulty
with credentials is that when the jury hears contradictory testimony from two
experts who seem to have even credentials, these credentials cannot yield even
prima facie weight. Coady argues that we can endow credentials with prima
facie weight by stamping them with the imprimatur of a leading university or
learned society. Three problems remain for this suggestion. First, it brings back
the related regress and question-begging problems: Being a famous university
in general is not the same as being a leading university in a particular field
(the University of California at San Francisco may not be a "leading
university," but, we are told, it is a leading center for medical research). We
can construct an analogy of proportion-UCSF: medical research :: University
of Pittsburgh: philosophical research. Indeed, many such examples can be

345. Id. at 286.

346. Id. at 287 (emphasisadded).
347. I considereda very similar question in connection with Daubert, see supra Section LB, and at
the beginning of this Article, see supra note 1 and accompanyingtext.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1633

found. In order for the nonexpert to find out which institutions were "leading"
in the proper way, she would have to find an expert already endorsed by one,
thus either begging the question or regressing. Another problem lingers even
after the credentialist solution. It is not clear how that purported solution can
handle cases of intra-disciplinary competition of credentialed experts. Such
competition occurs when two sets of witnesses who purport to be expert in the
same area (e.g., statistics, economics, genetics, epidemiology) and whose
credentials are evenly matched testify to contradictory or inconsistent
propositions. Nor is it clear how credentialism can handle cases of extra-
disciplinary competition of credentialed experts. This type of competition
occurs when two sets of witnesses who do not purport to be expert in the same
area, but whose credentials are evenly matched, offer testimony that is in some
way mutually undermining. Kenny provides a charming example, arguing that
the concept of the "irresistible impulse," brought into the criminal law by
expert psychiatric testimony, is philosophically incoherent and thus unworthy
of any serious credence in a courtroom:

The only remedy for this state of affairs will presumably be for the
prosecution to call a philosopher to testify that there cannot be any
such thing as an irresistible impulse, and therefore the accused cannot
have acted on one, any more than he can have murdered a married
bachelor or stolen a square circle. The desperate nature of this
proposal will, I hope, bring home vividly the indefensibility of the
present state of the law.348

And if Kenny, as credentialed as he is in the academic world, can lob such

extra-disciplinary firebombs at psychiatry, imagine what the inarguably well-
credentialed Paul Feyerabends and Richard Rortys of the world could do to
challenge the KJB-producing capacities of physics, biology, and mathematics,
were their expert testimony admitted in a trial. Extra-disciplinary challengers
like these well-credentialed scholars might not be correct on the merits, but
one must wonder how a nonexpert, using the credentials of academic prestige
as her guide, could make a justified judgment to reject their arguments. (This
kind of competitive testimony would, for the most part, have to be implied,
since such testimony is rarely admitted. But is it justifiably excluded by a
nonexpert judge?)
We reach a dialectical impasse, an aporia-our escape blocked by
opposing educated intuitions. On one side is the intuition that credentials must
be sufficient to enable KJB to arise from practical epistemic deference. On the
other side is the (to my mind) equally strong philosophical "intuition" that
without epistemic competence, credentials quickly crumble under scrutiny into
a pile of regressive, question-begging, underdeterminative heaps of unjustified

348. Kenny, supra note 64, at 56.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1634 The Yale Law Journal [Vol. 107: 1535

belief. The way through this impasse, I suggest, is by studying with meticulous
care, and rationally reconstructing, the step-by-step reasoning process a
nonexpert judge or juror must use in order to assess expert scientific testimony
in the course of reaching a conclusion in a legal dispute. As we shall see, the
inference process known as abduction plays a critical role in that reasoning
process. And it is some crucial features of abductive inference that will lead
us out of the impasse I have identified. I turn, then, to present a model of the
reasoning process the nonexpert uses to assess expert testimony, explaining the
role of abduction along the way.

VI. BREAKING THROUGHTHE DIALECTICALIMPASSE:A MODEL OF

RATIONALEPISTEMICDEFERENCETO EXPERTS

I shall begin with some preliminaries. As I discussed above,349I assume

that there is a useful distinction to be made between practical and theoretical
reasoning and between the practical and theoretical points of view from which
that reasoning proceeds. In light of that assumption, I argue that there are two
distinct kinds of "practical priority" involved in practical epistemic deference.
Relying on a simple model of the practical syllogism, I refer to one kind of
practical priority as "major premise" practical priority and to the other as
"minor premise" practical priority.
Major premise practical priority is a direct result of-perhaps even just
another way of expressing-the familiar view that one cannot "derive an ought
from an is," that in any valid practical syllogism, the set of propositions from
which a prescriptive conclusion is inferred must contain at least one
prescriptive premise that logically subsumes the subject of the prescriptive
conclusion. Minor premise practical priority is less noticed, though not
deservedly so. When a nonexpert practical reasoner solicits the view of a
theoretical expert in the course of making a practical decision, the minor
premise of the practical syllogism is actually the conclusion of a distinct
syllogism; it is the conclusion of an independent chain of reasoning by the
practical reasoner about the theoretical reasoning offered by the expert, which
functions as a lemma for the principal syllogism.

A. The Concept of "Practical Priority"

There are necessary limits to practical epistemic deference, even when the
nonexpert expressly seeks the guidance of the theoretical expert. The
explanation of "practical priority" in two premises of the standard practical
syllogism will bring out two of the most significant limits. According to a
standard model, adequate for my purposes here, the practical syllogism

349. See supra Section II.C.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1998] Scientific Expert Testimony 1635

contains two types of proposition: prescriptive propositions, and "descriptive"

or "theoretical" ones. Thus, the standard pattern for practical reasoning has the
following form:

Major premise: universal prescriptive proposition.

Minor premise: singular theoretical proposition.
Conclusion: singular prescriptive proposition.350

Slightly less abstract, a typical "legal syllogism" has this idealized form:

(1) All thieves ought to be hanged.

(2) Jones is a thief.
Therefore,
(3) Jones ought to be hanged.35'

I am interested in practical reasoning in cases in which the second premise

of the practical syllogism is a theoretical proposition that is capable of being
established by expert theoretical methods, such as those of science.
"Practical priority" refers to a simple feature of the practical syllogism,
that the major premise of any practical syllogism is a prescriptive practical
proposition. The major practical premise is "prior" to the minor theoretical
premise in that it determines the type of theoretical proposition that is relevant
for the purposes of making a practical judgment. The story could of course
become more complex if one sought to reflect in the schema for the practical
syllogism the possible occurrence of "thick concepts" like theft, courage,
murder, and the like.352 Philosophers most often recognize one type of

350. The analysis of practicalreasoningcould become more complicatedthan the simple syllogism
in the text reflectswere one to reflectin the model of practicalreasoningthe complete structureof singular
propositions,Classical categoricalsyllogistic logic calls for special treatmentof singularpropositions(e.g.,
"Socratesis a man").On the one hand,as ImmanuelKantobserved,"Logiciansarejustifiedin saying that,
in the employment of judgments in syllogisms, singularjudgments can be treated like those that are
universal."IMMANUELKANT, CRITIQUEOF PURE REASON 107 (NormanKemp Smith trans., Macmillan
1965) (1781). On the other hand, such a treatment,while reflecting the universality of the singular
proposition (by treating the subject term as a unit class), runs the risk of overlooking the singular
proposition's existential import (i.e., the proposition'sassertion that the unit class is not empty). One
solution is to treatthe singularpropositionas the conjunctionof a universalaffirmative(e.g., "All men are
mortal") and a particular affirmative (e.g., "Some men are mortal"). See IRVINGM. CoPI & CARLCOHEN,
INTRODUCTION TO LOGIC 412-14 (9th ed. 1994).
Note also that,accordingto manyphilosophers,defeasibilityis a featureof every practicalsyllogism.
Although defeasibility is not reflected in the simple practicalsyllogism above, I do account for it in the
model of practicalepistemicdeferencepresentedbelow. See, e.g., infraSubsectionVI.C.3 (describingsteps
(7) and (8) in my model).
351. For discussionof this form of the practicalsyllogism (using an example very similarto this one),
see GEORGHENRIK VON WRIGHT, THE VARIETIES OF GOODNESS162 (1963). Other forms of the practical
syllogism are possible. See, e.g., ROBERTAUDI, PRACTICALREASONING17-24 (1989); RAZ,PRACTICAL
REASON, supra note 132, at 17-24; VON WRIGHT,supra, at 160-71. I believe, however, that this one will
suffice for my purposes here.
352. For example, in the syllogism above, premise (2) might be considered either theoretical or
prescriptive, depending on one's account of thick practical concepts. I am inclined to think that thick
practical propositions, like (2), contain two separable components, one prescriptive and one theoretical. That

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1636 The Yale Law Journal [Vol. 107: 1535

practical priority-what I refer to here as "major premise" practical

priority-when they endorse the view that one cannot make a valid inference
to a particular prescriptive proposition unless somewhere in the set of premises
there appears a prescriptive proposition that logically subsumes the particular.
The claim of major premise practical priority really amounts to no more than
the much discussed claims that one cannot "derive an ought from an is" and
that there is a fundamental metaphysical distinction between fact and value,
and I do not dwell on it here. Instead, I turn to the feature of the basic
practical syllogism that has not been much remarked, but is of vital importance
for understanding practical-theoretical deference.

B. "Practical Priority" in the Minor Premise of a Practical Syllogism

The foregoing section discussed practical priority in the major premise.

There is yet another location of practical priority in practical reasoning. I refer
to it as "practical priority in the minor premise," but that is only partially
accurate. The fuller and more accurate claim is that when a practical reasoner
deploys a practical syllogism, the minor premise necessarily serves as both the
minor premise of one argument and the conclusion of a distinct argument. I
shall refer to these arguments as "primary"and "secondary," respectively.
Consider the following argument, whose three steps constitute what I am
calling a "primary" argument:

(1) This jurisdiction must convict all and only the persons who
commit crime X (major premise).353
(2) Jones committed crime X (minor premise).
Therefore,
(3) This jurisdiction must convict Jones.354

debate,however,is beyondthe scope of this Article.For discussionof thick ethical concepts,see BERNARD
WILLIAMS, ETHICS AND THE LIMITS OF PHILOSOPHY128-31 (1985). With obvious minor changes,
Williams's account can be extended beyond the realm of ethics to that of practicalreasoninggenerally.
353. Obviously this is highly idealized. It is possible that there is no actual legal system that even
aspires to this degree of accuracy,much less one that achieves it. It is accepted,and probablyinevitable,
that rules of criminalsubstanceand procedureare both overinclusiveand underinclusive,as measuredby
their background justifications. See P.S. ATIYAH & ROBERTS. SUMMERS,FORM AND SUBSTANCEIN
ANGLO-AMERICANLAW: A COMPARATIVE STUDY OF LEGALREASONING,LEGALTHEORY,AND LEGAL
INSTITUTIONS70-95 (1987); FREDERICKSCHAUER, PLAYING BY THE RULES: A PHILOSOPHICAL
EXAMINATION OF RULE-BASEDDECISION-MAKING IN LAWAND IN LIFE31-34 (1991). In the United States,
when the crime in question is a capital crime, there is under Supreme Court doctrine a demand for
"heightenedreliability"in the process of assessing guilt and makingthe decision to mete out the "ultimate
punishment."But even heightenedreliabilityis not thoughtby advocatesof capitalpunishmentto require
absolute reliability in determiningwho actually committed a capital crime. Some of these advocates
acknowledgethatsome mistakeswill be made, with the resultthatthe state will execute (and has executed)
some innocentpeople. But, these advocatesargue,overinclusivenessand underinclusivenessis a necessary
featureof all rules, and capitalpunishmentis desirableenough that we should toleratethis slippage. Some
even argue for an analogy: The innocentexecuted are like civilian casualties in a "justwar."
354. I assume for the sake of argumenteither that the predicateof the crime (e.g., "murder")is not
a thick practical concept, or, if it is one, that it can be broken down into prescriptiveand descriptive

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1998] Scientific Expert Testimony 1637

The warrant for premise (1) is practical, since it states a rule that prescribes
conduct, in this case, the conduct of the legal system in a given jurisdiction.
A proposition like (3) cannot be derived by valid rules of inference from a
proposition like (2) alone (at least when construing (2) as a descriptive
proposition355);some prescriptive proposition like (1) is needed to make the
inference to (3) valid. In this sense, (1), the major premise of this practical
syllogism, is logically prior to (2) in the resulting valid syllogism (although of
course the placement of (1) and (2) vis-A-vis each other in the set of premises
from which (3) is inferred is a matter of logical indifference). I have referred
to this phenomenon of the logic of practical reasoning as "major premise
practical priority," in which the practical major premise is logically prior to the
minor theoretical (descriptive) premise.
Now I want to focus on premise (2). I argue that when a theoretical
premise enters a practical syllogism of the sort previously illustrated (let us
call this the "primary argument"), it is always itself the conclusion of an
independent argument (let us call this the "secondary argument"). That is, (2)
serves double duty in a pair of nested syllogisms. At the "top" of that
independent, secondary argument is an additional practical premise. Put the
same point another way: Every theoretical premise in a primary practical
syllogism is, as it were, the conclusion of a secondary syllogism in which
there is also practical priority. This is the phenomenon I will refer to as "minor
premise practical priority."
Whether or not the practical reasoner is seeking to solicit instruction
from-and to defer epistemically to-a theoretical expert, this separate
syllogism is always enthymematically presupposed by the overall practical
syllogism. In most discussions of the practical syllogism, this separate
syllogism is left wholly enthymematic.356It is the separate argument that
requires several distinct but closely related judgments that the practical
reasoner must make in order to endorse as true the theoretical, descriptive
judgment that will serve as the minor premise of the primary syllogism. These
distinct judgments are reflected in the premises of the secondary argument,
which concludes with the theoretical judgment. Among these closely related
judgments reflected in the secondary argument, two are especially significant:

components(includinga descriptivecomponentlike "killedthe victim"and a prescriptivecomponentlike

"killed the victim without adequatejustification"),and that the theoreticaljudgment is a judgment about
whetherthe defendant'saction satisfied the criteriaof the descriptivepredicate.
355. See supra note 352 and accompanyingtext (discussinghow I would handlethe kinds of "thick"
practicalconcept-termsthat occur in many authoritativelegal norms,such as "murder,""theft,""assault,"
and "negligence").
356. Extending tradition, we might call it an enthymeme of the "fourth order." Some writers
distinguish "orders"of enthymemeaccordingto which line of the syllogism is unexpressed:"first-order"
if the majorpremise is unexpressed,"second-order"if the minor,and "third-order" if the conclusion. CoPI
& COHEN, supra note 350, at 294-95. I suggest "enthymeme of the fourth order" to reflect the fact that in
most rational reconstructions of the practical syllogism, the only step in the argument that is included is
the conclusion of the secondary argument.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1638 The Yale Law Journal [Vol. 107: 1535

a judgment about the level of epistemic confidence the practical reasoner

requires for her ultimate practical purposes (the purposes that guide the
primary argument); and a judgment about whether and how the epistemic
method on which the practical reasoner has chosen to rely (e.g., her own
perception, memory and inference, or expert testimony) meets the required
level of confidence.
This is quite abstract, and I shall explain it with several examples. I shall
begin by pointing out that the judgment that a practical reasoner must make
about the proper level of confidence required to arrive at the conclusion of the
secondary argument, which then serves as the minor premise of the primary
practical syllogism, is familiar in legal systems. Indeed, such systems have
elaborate evidentiary and procedural rules to guide legal decisionmakers in
making such judgments. The rules of what I have referred to as the "law's
epistemology" (rules governing admissibility, relevance, materiality, testimony,
documentation, hearsay, and judicial notice),357 as exemplified in Brown358
and Daubert,359are concerned with guiding judges and juries, and guiding
judges in the guiding of juries and other factfinders, in making these decisions.
In effect, these rules require the nonexpert practical legal reasoner to
proceduralize his caution about any theoretical claim about the world,
including any scientific claim, that is offered as a "candidate" for the minor
premise of the primary practical syllogism. Guided by the norms of legal
epistemology, the practical reasoner recognizes a practical obligation to be
cautious about scientific claims (indeed, about all expert claims). The caution
mandated by these norms is, as it were, institutionalized by procedural rules
that operate as presumptions against accepting scientific (and other) theoretical
claims in the course of applying a law to a person.
One might sum up the suggestion I am making here about the
institutionalized and proceduralized caution with which a practical reasoner
approaches theoretical claims as follows: What the facts are from a legal point
of view is not necessarily what the facts are from a scientific point of
view.360 This is the basic idea of "minor premise practical priority." The idea
that one should view facts as relative to the point of view in which facts are
assessed might seem to reinforce or at least cohere well with arguments, which
have some currency in both philosophical and legal scholarship, that "facts are
purpose-relative."36'With some such idea in mind, one might be inclined to
distinguish "scientific facts," facts that are relevant for scientific purposes,
from "legal facts" (or "practical facts"), facts that are relevant for legal

357. See supra Section I.A.

358. Brown v. Board of Educ., 347 U.S. 483 (1954); see supra Section I.C.
359. Daubertv. MerrellDow Pharms.,Inc., 509 U.S. 579 (1993); see supra Section I.B.
360. Again, this is a familiarfeatureof legal systems in which it is perfectlycoherentto say both that
Jones actually committedcriminalact X but is not legally guilty of having committedact X.
361. See, e.g., HILARYPUTNAM,REASON,TRUTHAND HISTORY127-49 (1981).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1998] Scientific Expert Testimony 1639

purposes (or other practical purposes). But here we must be careful, for some
of the "relativistic" theories, to which my own claim here might be likened,
are quite metaphysically suspect. Three more precise claims might be advanced
by the vaguer assertion that "facts are purpose-relative": (1) different facts are
salient from different points of view; (2) facts are actually different from
different points of view; and (3) from different points of view, different levels
of confidence (and possibly also different standards of epistemic appraisal) are
required to conclude that a putative fact is an actual fact.
Claim (1) seems uncontroversial and easily supported. I endorse it. Claim
(2) is fatally metaphysically suspect; it is far too idealist for comfort. Claim (3)
does not go nearly as far as (2) in the direction of questionable metaphysics,
but differs from and makes an epistemically more informative statement than
(1). It is all I will commit myself to here. Thus, taking advantage of the "point
of view" framework established earlier,362 it is in the sense expressed by (3)
that one should understand my point about the proceduralized caution of the
practical legal reasoner and my attendant claim that what the facts are from a
legal point of view is not necessarily what the facts are from a scientific point
of view.
Some legal reasoners are well aware, and as a whole there is in the legal
system increasing awareness, of the nature and importance of these different
points of view. A cogent articulation of this point is found, for example, in
DeLuca v. Merrell Dow Pharmaceuticals, Inc.363That case, decided by the
Third Circuit, dealt with the same issue treated in Daubert, namely whether the
"morning-after" drug Bendectin could have caused the plaintiff's birth defects.
The Third Circuit reversed a trial court judgment that expert testimony by a
well-credentialed epidemiologist was not admissible under the Federal Rules
of Evidence. Having discussed some of the basics of epidemiological analysis,
including the role of the "null hypothesis" in such research,3Mthe court
stated that on remand, the district court judge must keep in mind that the "fact
that a scientific community may require a particular level of assurance for its
own purposes before it will regard a null hypothesis as disproven does not
necessarily mean that expert opinion with somewhat less assurance is not
sufficiently reliable to be helpful in the context of civil litigation."365
Highlighting the difference between the level of epistemic confidence
required from a legal point of view and that required from a scientific point

362. See supra Sections II.B-C.

363. 911 F.2d 941 (3d Cir. 1990).
364. See id. at 945-49. Epidemiologistsuse the "null hypothesis"as a device for testing the accuracy
of a statisticalcorrelationthat has been discovered between some harmand a substancethat might have
caused the harm. The null hypothesis is the hypothesis that-contrary to some prima facie statistical
finding-there is actually no correlationbetween the two variables (harm caused and what might have
caused it). The techniqueis to assume arguendothat the "null hypothesis"is true, and then try to falsify
it.
365. Id. (emphasisadded).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1640 The Yale Law Journal [Vol. 107: 1535

of view is an important part of my overall analysis, and I return to it

below.366For the moment, I want to analyze the impact that this difference
makes on the structure of reasoning that a practical reasoner deploys when
soliciting expert scientific information from a theoretical expert. For heuristic
purposes, I will move toward presenting a complete model of this reasoning
process (doing the kind of philosophical analysis sometimes referred to as
"rational reconstruction") in slow steps.
Before considering an example of practical epistemic deference in a legal
setting, I begin with a somewhat simpler example of the kind of practical
moral reasoning in which the reasoner does not solicit expert information to
make the theoretical judgment that constitutes the minor premise of the
(primary) practical syllogism, but instead relies on her own epistemic
competence to make that judgment.367Assume that the following is a sound
syllogism (a valid inference with true premises) in which Jones (the referent
of 'I' in premise (2)) reasons about his moral obligation to Williams:

(1) Every person always ought to keep his promises.

(2) I [Jones] promised to meet Williams for lunch.
Therefore,
(3) I ought to meet Williams for lunch.

This simple syllogism is a bit of practical reasoning about what is required

of Jones from a moral point of view. The syllogism Jones constructs consists
of two prescriptive propositions, (1) and (3), and a theoretical proposition, (2).
I claim that (2) must be understood as the conclusion of an argument that is
independent of the syllogism in which it serves as the minor premise.
Proposition (2) rests on (is warranted by) a chain of propositions of which it
is the conclusion. Jones, the practical reasoner, is warranted in asserting (2) in
part because he ought to believe that (2) is true. The "ought" in (2) is an
epistemic one,368which seeks to connect the practical norm expressed in (1)
to some feature of the world (that is, to what it is warranted to believe about
this feature of the world), namely, that Jones actually performed the actions
that constitute promising. By contrast to the two "oughts" in propositions (1)
and (3), both of which are practical (and in this case, moral), the minor
premise connects the moral norm in (1) to the world.
I maintain that the proposition that Jones ought (epistemically) to believe
proposition (2) is itself an inference from further propositions, such as that the
belief stated in this propositionwas producedby a sufficientlywarranting

366. See infra Section VI.C.

367. Of course, moral reasonerscan and often should solicit expert informationin order to make
sufficiently justified moral decisions. For a powerful and elegantly concise suggestion along these lines,
see Peter Singer, Moral Experts,32 ANALYSIS115 (1971-1972).
368. We may say that the reasonerought to believe (2) because it is a warrantedbelief.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific ExpertTestimony 1641

epistemic procedure.Perhapsthis procedureconsists in Jones's consultinghis

knowledge of the language of promising, his senses, and his memory for
events that surroundedhis act of promising.(I am assumingthatJones, unlike
a thoroughgoingskeptic,believes the evidence of his own senses and memory
and everyday knowledge, and that he does not treat this practicaljudgment
about Jones's obligations as one that calls for expert evidence.) Again, the
judgment of sufficient warrantis an epistemic one; Smith relies on the
"testimony"of his memory and senses and everydayknowledge because he
believes he has good reason to rely on them as a procedurefor producing
truthsabout the world.
If I am right, what this simple example reveals is that the minorpremise
of a practical syllogism, which is a theoretical proposition, rests
enthymematicallyon a series of related epistemic judgments. Given the
putativerelationbetweentheoreticaljudgmentsand facts aboutthe world, this
is not surprising.But a while back I began this argumentby claiming that
there was practical priority even in the minor, theoretical premise of a
practical syllogism. So far, we have seen only additional theoretical
propositionssupportingit. Where is the "practicallyprior"premise?
The answer,I believe, is in the practicalreasoner'sjudgmentof sufficient
warrant.Recall that one of the underlyingjudgmentsthat warrantedJones's
endorsement of minor premise (2) (namely, that Jones promised to meet
Williamsfor lunch)was an additionaljudgmentthatJones oughtepistemically
to believe that the minorpremise was true, because he also believed that the
judgment in the minor premise was producedby a sufficiently warranting
epistemic procedure.But the phrase "sufficientwarrant"fails to capturethe
full judgment that Jones must make. More precisely, he must decide what
quantity of warrantor what quality of warrant (if it does not admit of
quantification)he needs for practical purposes.369 To explain this point, I
call again on Laudan'saxiological model of a rationalenterprise,370 here the
"enterprise"of moral reasoning.The moralreasonerhas certainoverall aims
(which might be explainedin deontologicalor teleologicalterms)and deploys
methods to serve those aims, which in turn produce particular moral
judgments. Consider again the simple syllogism in the example above, in
which Jones reasons about how he ought, from a moral point of view, to
behave towardWilliams.Let us supposethatJones seeks to make surethathe
keeps his promise to Williams. One of the methods Jones might choose to
deploy to help bring about that end is to consult his own memoryregarding
the exact content of that promise. Jones would be deciding not just that the
testimony of his memory and senses and everyday knowledge was
"warranted,"but that it was sufficientlywarrantedfor his moral (practical)

369. See supra Section II.C (discussing quantityand qualityof warrant).

370. See supra notes 143-158 and accompanyingtext.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1642 The Yale Law Journal [Vol. 107: 1535

purposes. Using that method, he could produce the individual judgment that
he ought to meet Williams for lunch. Given a different practical task, governed
and guided by different or additional practical norms, he might well have
concluded that this kind of testimony was not adequate for his purposes.
Consider now a different practical syllogism:

(1) One always ought to assassinate a politician whose continued

life is a serious threat to the political liberty of some nation(s).
(2) Politician H's continued life is a threat to the political liberty
of some nation(s).
Therefore,
(3) One ought to assassinate politician H.

As with every practical syllogism, the major premise establishes criteria whose
satisfaction in a particular case (for a particular politician) it is the job of the
minor premise to ascertain; that is, it is up to the minor premise to feed into
the practical decision some relevant information about the world (information
made relevant by the major premise). Surely for such criteria as the practical
norm (1) establishes, a practical reasoner would want a better warranted
theoretical test than he would for a decision about whether to keep a promise
to meet a friend for lunch.37' That is, although the theoretical "testimony" of
everyday knowledge, the senses, and memory might be sufficiently warranted
for some practical purposes, it will not be adequate for every practical purpose.
Because every practical reasoner must decide the quantity or quality of warrant
that is practically necessary for his decision, even his assertion of a theoretical
premise depends on a practical decision about the quantity or quality of
warrant he needs. That is the feature of practical reasoning I am calling "minor
premise practical priority."The procedural mechanisms in the American courts,
some of them constitutionally grounded, exemplify minor premise practical
priority in the different "burdens of persuasion" that such rules establish for
criminal prosecution (the state must prove guilt "beyond a reasonable
doubt"372)and for civil suit (usually, the plaintiff must prove his claim by
a "preponderance of the evidence"373).
Presented schematically (and in the first person, though obviously it could
be easily adapted to second- or third-person assessments of practical duty), the
practical syllogism about promise keeping looks like this:

371. I am assuming, of course, a moral reasonerwhose moralitycould countenancekilling for this

reason-presumably a judgmentfar more congenial to teleological thanto deontologicalsystems. It is also
possible that a utilitariancalculatormight well factor in even an admittedlysmall chance of achieving a
hugely importantoutcome, and allow such a calculation to lower the level of proof of (2) she requires
before reachingthe conclusion in (3).
372. See In re Winship,397 U.S. 358, 364 (1970).
373. See, e.g., CAL. EVID. CODE ? 115 (West 1995) ("Exceptas otherwiseprovidedby law, the burden
of proof requiresproof by a preponderanceof the evidence.").

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1643

(1) I always ought to keep my promises.

(2c) The "testimony" of my everyday knowledge of
promising behavior, my senses, and my memory is sufficiently
warranting for the purposes of guiding my promise-keeping behavior.
(2b) The judgment that I promised to meet my friend for
lunch is the result of a sufficiently warranting process (namely, the
"testimony" of my everyday knowledge of promising behavior, my
senses, and my memory).
(2a) I ought (epistemically) to believe that I promised to
meet my friend for lunch.
(2) I promised to meet my friend for lunch.
Therefore,
(3) I ought (morally) to meet my friend for lunch.

The indented propositions (2c) to (2a) represent the independent argument of

which (2) is the conclusion. While (2b), (2a), and (2) are theoretical
propositions, governed by epistemic rather than practical norms, (2c)-the
judgment that a given theoretical-judgment-producing process is sufficiently
warranting for practical purposes-is a practical proposition, governed by
practical norms. Every minor premise in every practical syllogism rests,
ultimately, on some practical norm like (2c), which governs the independent
secondary argument that feeds into the primary argument consisting of
propositions (1), (2), and (3). It is in this way that the two intersecting
arguments exhibit minor premise practical priority.
I turn now to the more complicated case of practical reasoning by a
nonexpert who solicits expert testimony to help reach a practical judgment that
is the conclusion of practical reasoning. My goal here is to show how minor
premise practical priority obtains in such cases as well. Consider the following
very abstract syllogistic representation of a definition of a legal concept of a
type quite common in statutory, administrative, and common law:

(1) All persons who have performed actions 'D have committed
actions 0.
(2) Williams performed actions (D.
Therefore,
(3) Williams has committed actions 0.

Propositions like those represented by the schema in (1) perform two "speech
act" functions. One is to offer stipulative definitional criteria (necessary or
sufficient conditions) for legal concepts; these stipulated criterial propositions
have the form, for actions (D and legal concepts 0, of "All instances of ( are
0," or "To be an instance of 0, an act must be (D," or "An act is an instance
of 0 if and only if it is an instance of (." Consider, for example, the
following criterial proposition, drawn from a state statute, which offers a
stipulative sufficient condition (i.e., the linguistic predicate for the actions ()

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1644 The Yale Law Journal [Vol. 107: 1535

for "Driving with Excess Blood Alcohol content" (i.e., the linguistic predicate
for the legal concept 0): "It is unlawful and punishable as provided in ? 28-
692.01 for any person to drive or be in actual physical control of any vehicle
within this state while there is 0.10 or more alcohol concentration in the
person's blood or breath at the time of the alleged offense."374The second
"speech act" function of these stipulative criterial definitions is to give the
definition force of law; the statute begins with the words 'It is unlawful,'
words which have the same effect in context as if the statute had begun, 'It is
hereby declared unlawful.' Usually such definitions are accompanied by some
statement of or reference to the legal effect of performing actions 4; in the
proposition above, note the phrase "punishable as provided in ? 28-692.01."
Although the example above is from a criminal statute (which has since been
amended), the same basic framework of stipulated criterial definitions
mandated with force of law explains definitions in civil law. It also explains
such definitions as they are promulgated in administrative regulations and in
judicial decisions.
Stipulative criterial definitions like proposition (1) are practical legal
propositions375that link the performance of actions 'D to the commission of
actions described by legal concept &. They also serve as major premises in the
practical syllogisms that legal reasoners use to apply the defined legal concepts
to individual cases. My special concern is with instances in which the
nonexpert practical reasoner solicits evidence from a scientific expert in order
to adopt and endorse a minor premise in the practical syllogism. Continuing
the example of the stipulated definition of "Driving with Excess Blood Alcohol
Content," the syllogism fashioned by the practical nonexpert judge or jury
considering a case in which Williams, a defendant, was charged under this
statutory proposition is as follows:

Major Premise: (1) All persons who drive or are in actual

physical control of any vehicle within the state while there is 0.10 or
more alcohol concentration in the person's blood or breath at the time
of the alleged offense are subject to [specified] legal sanction.
Minor Premise: (2) Williams drove or was in actual physical
control of a vehicle within the state while there was 0.10 or more
alcohol concentration in his blood or breath at the time of the alleged
offense.
Conclusion: (3) Williams is subject to [specified] legal sanction.

374. ARIZ. REv. STAT. ANN. ? 28-692(B) (West 1989) (amended 1993).
375. On some positivist views, these are not practical but theoretical propositions. See supra note 3.
My analysis of epistemic deference and its rational structure does not depend on accepting or rejecting such
views. If legal propositions are theoretical, then what I am calling "minor premise practical priority" could
be reframed as something like "minor premise priority."

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1645

It is common for a jury or judge that is applying the legal concept specified
in the major premise, (1), to call on expert scientific testimony in order to
establish the minor premise, (2). As in the case law applying the variety of
"driving under the influence" provisions that appear in criminal statutes clearly
reflects, there are many ways in which expert scientific evidence can be both
relevant and material to the outcome of a case. Experts might be called upon
to determine the blood alcohol content of an arrested driver at some time after
arrest, or to establish what his blood alcohol content was at the time of arrest
based on an extrapolation from his blood alcohol content at the time of testing
(which is often more than an hour after arrest). Testimony on these issues
might also involve testimony about the scientific reliability of the methods
used to test blood alcohol content or to "relate back" the content at the time
of testing to the content at the time of arrest.376
In a case in which the nonexpert judge or jury solicits expert scientific
information in order to apply a legal concept like "driving with Excess Blood
Alcohol content," the nonexpert reasoner is seeking expert advice in order to
decide which minor premise to endorse-i.e., to determine whether to accept
or reject the prosecution contention that

(2) [the defendant] performed actions (D.

When one reconstructs the nonexpert's syllogistic reasoning, one might

conclude that (2) is simply the statement made by the expert to the judge or
jury,378 but that would not be an accurate reconstruction of the nonexpert's

376. To takejust one of (at least) hundredsof possible case law examples, in State ex rel. McDougall
v. Albrecht, 811 P.2d 791 (Ariz. Ct. App. 1991), the court held that a defendant'sfailure to pass field
sobrietytests, coupled with his speeding and failureto stop at a red light, providedsufficient evidence for
the admission of his blood alcohol level at the time of testing, which was significantlyafter the time of
arrest.This evidence was admissible,held the court, because the evidence of speeding, of failure to stop,
and of failure to pass field sobrietytests constitutedsufficientevidence for the expert to "relate. . . back"
the blood alcohol content at the time of testing to that at the time of arrest.Id. at 795. The court also
rejected some of the expert testimony in this case. The expert used evidence of the field sobriety test,
coupled with the evidence of the post-arrestblood alcohol content,to concludethattherewas a 67% chance
that Williams had a blood alcohol content greaterthan 0.10%. See id. at 793. Rejecting the admissibility
of this particularcalculation,based solely on the post-arrestblood alcohol content and the failure to pass
field sobriety tests, the court stated that the expert's "characterizationof the field sobriety tests was
misleading in that the jury may have given the tests undue significanceas scientific truths,when, in fact,
they measureonly gross neurologicalimpairment."Id. at 795-96. The court continued:"Wehave concern
about the misleadinginfluence [that]such speculativeconclusions unsupportedby controlledexperiments
may have on a jury."Id. at 796 (footnote omitted).
377. As I have already mentioned,this analysis applies just as well to "thick legal concepts" like
murderand reckless endangermentas it does to blood alcohol in excess of 0.10. For thick legal concepts,
the scientific expert testifies to the truthor falsity of the descriptiveevidence.
378. As noted supra Section LB, rules of evidence divide decisionmakinglabor between two types
of practical reasoner-judges and juries. The rules allow the judge to decide which experts, and which
information,is relevantto a given practicaldecision; the jury then hears from those experts that pass that
test. The jury then listens to the screened experts and decides to what extent to credit the testimony in
determiningwhetherthe defendant'sconductsatisfiedthe predicatein question(e.g., "murder").Sometimes
the judge himself is the factfinder,and so the same personperformsboth tasks; while this is quite rarein

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1646 The Yale Law Journal [Vol. 107: 1535

full chain of reasoning. Although (2) is indeed a theoretical proposition, it is

a theoretical proposition made by the practical reasoner on the basis of expert
evidence. That is, (2) represents the conclusion of a chain of reasoning that
includes a set of related judgments about the reasoning that the expert himself
used to arrive at the judgment he reports testimonially to the practical reasoner.
To reconstruct that chain in a plausible manner, we might retroduct back from
(2) itself (the conclusion of the "secondary" argument, distinct from the
primary argument in which (2) is the minor premise) to those implicit
judgments that would, from the practical reasoner's point of view, warrant the
practical reasoner's assertion of (2).
Underlying (that is, offering warrant for) proposition (2) is some
proposition to the effect that "we (the practical reasoners) ought (epistemically)
to believe that Williams committed actions e." Underlying that proposition,
in turn, is a set of propositions to the following effect:

The judgment that Williams committed actions E is the conclusion of

an expert-scientific process (call it P), and we ought to convict on the
basis of the results of P, because P produces sufficiently warranted
beliefs, and the law ought to convict on the basis of judgments
yielded by expert-scientific processes that produce such beliefs
because acting on them is an acceptable way to achieve the practical
goal of treating all persons who have committed actions in accord
with the legal sanctions (or legal benefits)379that the law specifies
for those who commit actions.

C. A Simplified Model of Practical Epistemic Deference

We may present this reasoning more schematically as an inferential chain

that presents the proposition that appears in the minor premise in the primary
syllogism (i.e., proposition (2)) as the conclusion of a distinct secondary
argument whose conclusion is (2) itself. I now present the abstract form of this
distinct inferential chain, with some comments on the source of each of the
premises. What I present in this section is only a preliminary version of the
model of this reasoning process-to be made more complex, and more
explanatorily adequate below.380I begin with a wholly abstract statement of
the "secondary" argument that a nonexpert practical reasoner must use in
deferring to a scientific expert. Then I offer detailed step-by-step analysis.

criminalcases, it is less so in civil ones. Thatdifferentpersonsperformthese tasks is not importantfor my

purposeshere.
379. This model of the practicallegal reasoner'sreasoningapplies equally to criminalsanctions and
to the benefits partiescan obtain from the "enabling"rules of contract,property,tax law, etc.
380. See infra Sections VI.D-E.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1998] Scientific Expert Testimony 1647

1. Summary Presentation of the Model

(1) If an expert's scientific process (P) produces a judgment (I)

that rises to the level of confidence of the appropriate standard of
epistemic appraisal specified by the practical reasoner (call the
specified level L), then the judgment that P produces is true for given
practical purposes (call these purposes R).
(2) If j is true for R, then if a practical reasoner has R (and has
no other inconsistent practical purposes), then that reasoner ought to
infer ].
(3) If a practical reasoner ought to infer j, then j.
(4) P produces judgments that satisfy L.
(5) j is the result of P.
(6) j is true for R.
(7) If a practical reasoner has R (and has no other inconsistent
practical purposes), then that reasoner ought to infer j.
(8) The practical reasoner has R (and has no other inconsistent
practical purposes).
(9) The practical reasoner ought to inferj.
Therefore,
(10) j.

2. Comment on Step (1)

If an expert's scientific process (P) produces a judgment (call itj) that

rises to the level of confidence of the appropriate standard of
epistemic appraisal specified by the practical reasoner (call the
specified level L), then the judgment (call it j) that P produces is true
for given practical purposes (call those purposes R).

Step (1) consists of the practical judgment, made by the deferring

nonexpert practical reasoner, about which expert's use of scientific processes
produces judgments that are epistemically satisfactory for the practical
purposes at hand.38' The reference to the expert's scientific process (rather
than simply to expert scientific processes) reflects the fact that it is the expert's

381. To anchorthis abstractionwith an example, recall the case of McDougall, in which the Arizona
criminal court ruled inadmissiblethe scientific expert's testimony that there was a 67% chance that the
defendanthad been driving with a 0.10% blood alcohol content. The court stated:
A courtroomis not a researchlaboratory.The fate of a defendantin a criminal prosecution
should not hang on its ability to successfully rebutscientificevidence which bears an "auraof
special reliabilityand trustworthiness,"although,in realitythe witness is testifyingon the basis
of an unprovedhypothesisin an isolated experimentwhich has yet to gain generalacceptance
in the field.
McDougall, 811 P.2d at 796 (quotingUnited States v. Brown, 557 F.2d 541, 556 (6th Cir. 1977)). In the
terms of my analysis, the Arizonacourt decided that the scientific methodthe expert had used to arriveat
the 67% figure did not produce judgments that were epistemically satisfactoryfor the practical legal
purposeof effecting a just conviction of a criminaldefendant.I discuss the normativeconstraintsinvolved
in this kind of decision in PartVII.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1648 The Yale Law Journal [Vol. 107: 1535

use of the process, and not the process itself, to which the nonexpert
defers.382One could, I suppose, consider the overall process to include both
the scientific techniques deployed and the person deploying them, but keeping
them separate seems to allow for more targeted diagnoses of failures and
successes.
There are two distinct terms of epistemic valuation that must be carefully
distinguished in (1). One is what I shall refer to as the level of confidence,
labeled by the variable L to indicate that different practical reasoners will set
different required levels of confidence. The other is what I shall refer to as the
standard of epistemic appraisal, by whose metric the level of confidence is
measured. Because, as I will explain, different standards (and attendant levels
of confidence) are possible, practical reasoners must also choose one or
another from among the possibilities (in a moment, I explain why I refer here
to practical reasoners and not just to one practical reasoner). In step (1), the
practical reasoners must make a decision about, and adopt, both a standard of
epistemic appraisal and a level of confidence that is measured by that standard.
I refer to practical reasoners as the decisionmakers in step (1) because
here, as in other areas of legal epistemology, we must be sensitive to the
division of labor among legislature, judge, and jury. The practical reasoners
principally responsible for setting the standard of epistemic appraisal and level
of confidence are legal officials, like legislatures and judges. These officials
reflect their choices about standards of appraisal in rules of evidence and
procedure, jury instructions, and other such devices. Note, however, that
different jurisdictions, different courts within a jurisdiction, or even different
judges within the same court, will make different decisions on the issue. It is
in that way that a "choice" must be made among distinct "possible" standards
of epistemic appraisal.
In what sense must these practical authorities "choose" from among
different possible standards of epistemic appraisal, and in what sense are
several such standards "possible"? Here yet another distinction is important,
a distinction between two senses in which the practical authority might make
a commitment to a standard of epistemic appraisal (and corresponding level of
confidence).
One is a "thick" commitment to a deep epistemological theory-that is, a
full-blown theoretical commitment to the truth of a particular epistemological
theory. Such commitments are clearly made by some evidence scholars and
some judges regarding the question of how best to explain the kinds of
epistemic judgments that judges and juries make when they evaluate evidence.
The leading debate on this issue among evidence jurists is whether a
factfinder's judgments are best explained in probabilistic terms or instead in

382. See supra Section III.A (explaining the distinction between believing a person and believing a
proposition).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1649

nonprobabilistic ones. The debate has tended to focus on the question of what
kind of reasoning process a juror or judge is using when she "weighs"
evidence and arrives at a belief on the basis of that weighing; is she
performing a probabilistic calculation (perhaps tacitly) or using some other
kind of rational cognitive processing? Evidential "probabilists" answer this
question in the affirmative.383This debate among evidence theorists parallels,
and often draws on, similar debates in general epistemology (where the
champions of probabilism are often found in the ranks of reliabilists)384and
philosophy of science. A "thick" theoretical commitment by a practical
reasoning authority to a probabilistic theory of evidential judgment can be
reflected in judicial and legislative rules of evidence and procedure, advisory
committee notes, and jury instructions, as well as in other ways. I shall play
out some of this debate with an example just below. First, however, I should
offer a word about the oddity or perhaps even the implausibility of claiming
that legal officials commit themselves and their jurisdictions to "thick"
epistemological theories concerning the nature of evidential assessment.
In this area of law, as in many others, there is an intellectual division of
labor between "high theory" workers and the appliers and administrators of
theory-motivated doctrines. The paradigm for this division of labor is the
common law itself. In that system, some judges, scholars, and lawyers take the
lead in organizing, systematizing, analyzing, rationalizing, and revising
doctrines and the theories that motivate them (e.g., theories of justice and
equity in contract, tort, constitutional, and property law; theories of mind and
motivation in the criminal law; economic theories throughout public and
private law). These are the Holmeses, the Cardozos, the Brandeises, the
Learned Hands, the Posners, and the Corbins, as well as innumerable scholars.
These jurists organize and reorganize whole lines of cases, propose values to
explain and criticize and motivate changes in common law doctrines. These
"high theory" jurists can indeed quite plausibly be seen as making thick
theoretical commitments to metaphysical and epistemological theories of the
sort seen in probabilistic accounts of factfinding judgments. But of course not
all jurists, and certainly not all judges, make or even attempt to make deep
theoretical commitments of this sort. Many defer epistemologically to their
high-theory brethren, administering doctrines articulated by the high theorists,
but without engaging in sophisticated high theory themselves. These judges
tend only to administer the doctrines that high theorists create for the law's
epistemology. They tend to make only the incremental changes that are
inevitable in the face of gaps, conflicts, and ambiguities that attend any system
of laws. Quite often they deploy the resources of analogy to make these
incremental changes, not least in using analogy-warranting rationales

383. Some of the classic argumentsin this debate are cited and discussed below. See infra note 385.
384. See, e.g., ALVIN I. GOLDMAN, EPISTEMOLOGY AND COGNITION (1986); NozICK, supra note 5.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1650 The Yale Law Journal [Vol. 107: 1535

developed by high-theory judges. And as it is in the common law, so it is in

the amalgam of common law, legislation, and administration that comprises the
official rules of legal epistemology.
I asserted that there are two senses in which legal officials can make a
commitment to an epistemological theory when they choose from among
different possible standpoints of epistemic appraisal. To complement the
"thick" sense of full-blown, high-level theoretical commitment, there is also
a "thin" sense in which the legal official is concerned, not so much with the
ultimate truth about belief (say, jurors' beliefs formed in assessing evidence),
but rather with the proper way explicitly to instruct jurors and judges to think
when they assess evidence. The thin account is not concerned with ultimate
truth but with an empirical judgment of the likely impact of different kinds of
instructions given to factfinders. Here again there are probabilist and anti-
probabilist camps. The former believe that factfinders ought to be told and
taught explicitly to use probabilistic methods in assessing evidence (typically,
they endorse Bayesian methods). To be sure many, perhaps even all, of these
probabilists are also committed to probabilism as a "thick epistemological"
account of the assessment of evidence, but there is no necessary connection
between thick and thin views. In the anti-probabilist camp, one might well find
theorists who believe that probabilism is the correct "thick" account of the
nature of evidential judgments by factfinders, but who also believe that it
would be a mistake for courts to adopt a "thin" evidentiary practice of
instructing factfinders to try consciously to assess evidence in probabilistic
terms. A thin doctrinal practice of that sort, they might argue, is likely to
create significantly greater factfinding errors than there would be when
factfinders did not try to "do probabilities" explicitly. (Analogously, a "cooking
theorist" might think a cook better advised to think in terms of "pinch,"
"dash," and "twist" than to try consciously to deploy more precise quantitative
measures, even though every pinch, dash and twist can be fairly precisely
measured.)
One should understand my own claim that the practical reasoners must
"choose" from among "possible" standards of epistemic appraisal as a claim
that is made at the meta-level. I am not endorsing a claim that there really are
several distinct standards of epistemic appraisal-it may well be that there is
actually only one. But even if one account, such as the "thick" probabilistic
account, turns out to be the truth about epistemic appraisal, it is also clear that
there exist competing theories about which standard best captures the process
of evidentially-based belief formation-that is, about which "thick" account is
correct. Moreover, as I. have just suggested, there are even competing views
about whether there is indeed only one correct standard of epistemic appraisal
to which all others are reducible, or whether instead there is a plurality of
standards. Again, my quite limited point here is that the "decision" that
practical reasoners must make in step (1) is either a commitment to a thick

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1998] Scientific Expert Testimony 1651

metaphysical and epistemological theory or a commitment to a thinner view

of how the factfinder ought to be instructed to think about evidence. Which of
these types of commitment the legal system and its coordinated decisionmakers
make in step (1) is not important for the analysis here; what is important is to
note that one or the other of these commitments must be made in the first step
of practical epistemic deference.
An example will surely help. Suppose the case presented to the practical
reasoner is a tort case for wrongful death. The case involves a fatal hit-and-run
vehicle collision on an enclosed high-security military base; an army officer
is killed in the collision. The case is brought in a state court. The plaintiff is
the wife of the deceased officer; the defendant another officer, who drives a
jeep. The state has typical rules of procedure and evidence; specifically, the
burden of proof it imposes in this (and most other) civil actions requires that
the plaintiff prove his case by a "preponderance of the evidence."
The high-security nature of the base is such that experts and their attorneys
can gather exact information about every vehicle that was present on the base
(in working order, etc.) at the time of the accident. An imprint of a kind of
steel bumper that is unique to a particular brand of military jeep is found on
the wrecked car driven by the deceased, and both sides concede that this
bumper imprint was caused by the offending vehicle. A forensic expert (trained
in the engineering of "accident reconstruction," for example) testifies that, as
the result of a study of all the jeeps on the base that contained this type of
bumper, it is 51% likely that the defendant's jeep, which is equipped with the
type of bumper that could have caused this imprint, caused the accident.385
The trial proceeds, and no other evidence is introduced (this is something of
a stylization, but it makes my point simpler to explain). Just before the jury
retires for deliberation, the court instructs that the "preponderance of the
evidence" standard governs this case, and explains this burden of proof as
follows:

385. I offer one version of an example familiarin the pro- and anti-probabilistliterature.In a seminal
article,LaurenceTribeoffers the "bluebus"hypotheticalto illustrateproblemswith probabilismin a civil
case. In this hypothetical,the plaintiff seeks recoveryafter being struckby a blue bus in a town in which
four-fifthsof the blue buses are operatedby the defendant.See LaurenceH. Tribe, Trialby Mathematics:
Precision and Ritual in the Legal Process, 84 HARv.L. REV. 1329, 1340-41 (1971). L. JonathanCohen
offers the "gate crasher"hypotheticalto raise some of the same issues. Suppose one thousandpeople are
seated in seats at the rodeo, but only 499 have paid. Payment was in cash, and there is no other proof
regardingwhetherany given individualpaid. Can the rodeo owner collect from (or win a civil or criminal
action against)a randomlychosen person?Can he collect againstevery personin the rodeo?Cohen argues
no. See L. JONATHAN COHEN, THE PROBABLE AND THE PROVABLE 75-81 (1977). CharlesNesson offers
a famous example in the criminalsetting. Suppose that in a prisonyard, thereis clear evidence that 24 out
of 25 prisonersparticipatedin the murderof a guard. There is no other evidence regardingany of the
individual prisoners. Can one prisoner of the 25, randomly chosen, be convicted of murder on this
evidence? Nesson argues no. See CharlesR. Nesson, ReasonableDoubt and Permissive Inferences: The
Valueof Complexity,92 HARV. L. REV. 1187, 1192-99 (1979).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1652 The Yale Law Journal [Vol. 107: 1535

The most acceptable meaning to be given to the expression, proof by

a preponderance, seems to be proof which leads a jury to find that the
existence of the contested fact is more probable than its nonexistence.
Thus the preponderance of the evidence is your belief in the
preponderance of probability, which can be restated as the burden of
showing odds of at least 51 to 49 that such and such has taken place
or is so.386

Relying on the expert testimony, the jury finds the defendant liable.
This court (I assume) both "thickly" and "thinly" endorses a subjective
probabilistic interpretation of the "preponderance of the evidence" burden of
proof.387 On this interpretation, the burden of proof is measured by a
probability that reflects not the "objective" relative frequency of the event (the
fatal crash with this particular defendant's vehicle) in repeated trials-the crash
is, after all, a unique event. Instead, according to the subjective probabilistic
interpretation, a factfinder's judgment that a given factual claim is supported
by a preponderance of the evidence reflects a rational juror's willingness to
bet, given even odds, that the event was more likely to have occurred than not.
In the terms I offered above, subjective probability is the standard of epistemic
appraisal that the court has directed the jury to use. The jury's specified level
of confidence is greater than 50%. This example thus illustrates my claim
about the exact types of epistemic "decision" a practical reasoner must make
in step (1) of the overall process of deferring epistemically to a scientific
expert. The nonexpert practical reasoner must decide both what standard of
epistemic appraisal to use and what specific level of confidence, as measured
by that standard, is required for his practical purposes. In my example, the
judge, perhaps guided by a legislative rule, decides to use a subjective
probabilistic standard with a level of confidence of greater than 50%.
A jurist's adherence to a probabilistic explanation of burdens of proof can
lead that jurist (judge, legislator, or scholar) to adopt a probabilistic standard
of epistemic appraisal. As a "thick" theoretical explanation of long-existing
practice with burdens of proof in legal systems, the probabilistic explanation
does indeed have significant explanatory virtues. One of those virtues is its
capacity elegantly to account for the different burdens of proof that courts (and
legislatures) impose on plaintiff or prosecutor for different kinds of action.388

386. I have amalgamated this example of a jury instruction from CHARLES T. MCCORMICK,
MCCORMICK'SHANDBOOKOF THE LAW OF EVIDENCE794 (Edward W. Cleary ed., 2d ed. 1972); and
Davies v. Taylor, 1974 App. Cas. 207, 219 (Eng.).
387. Some "thickly theoretically" inclined courts do indeed view the preponderance of the evidence
standard this way. One British court, for example, asserted that "the concept of proof on balance of
probabilities, which can be restated as the burden of showing odds of at least 51 to 49 that such and such
has taken place or is so." Davies, 1974 App. Cas. at 219.
388. An official commission comment on the California rules of evidence discusses different burdens
of proof in this way:
Usually, the burden of proof requires a party to convince the trier of fact that the existence of
a particular fact is more probable than its nonexistence-a degree of proof usually described

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
 1998] Scientific Expert Testimony 1653

It is perhaps worth emphasizing the importance of keeping distinct the burden

of proof a jurisdiction imposes on a litigant (preponderance of evidence, clear
and convincing, beyond a reasonable doubt) and the standard of epistemic
appraisal that a factfinder uses to assess a litigant's factual claims. The latter
is an interpretation and explanation of the former. Some jurists explain and
interpret the preponderance burden of proof in probabilistic terms, whether
subjective or objective. Others interpret that and other burdens in
nonprobabilistic ways.
Having argued that standards of epistemic appraisal are interpretations of
burdens of proof, I must also explain why the burdens of proof are not
themselves standards of epistemic appraisal in my sense. To be sure, it seems
likely that judges and legislatures originally adopted the familiar phrases
associated with different burdens of proof (preponderance, beyond a reasonable
doubt, etc.) as standards of epistemic appraisal, rather than as propositions that
can be interpreted by probabilistic or nonprobabilistic standards. But over the
course of the common and statutory law of evidence, these phrases have
become canonical and relatively independent of the original intentions of their
framers. It is by virtue of this relative independence of the canonical phrases
that we can and should see standards of epistemic appraisal, such as
probabilistic and nonprobabilistic standards, as interpretations of burdens of
proof.
I observed above that a probabilistic explanation of burdens of proof has
explanatory virtues that lead some jurists to recommend probabilistic standards
of epistemic appraisal. But not all jurists endorse probabilistic standards, and
even those who believe that the probabilistic "thick" theoretical explanation of
factfinders' judgments is correct can consistently decline to endorse a
probabilistic standard of epistemic appraisal as the best explanation of legal
burdens of proof. Moreover, not every court or legislature is committed, either
thickly or thinly, to this standard (and its attendant level-of-confidence
measure). To see how the "anti-probabilist" views are among the possible
choices an epistemically deferring practical reasoner can reasonably make,
recall the ongoing debate among Anglo-American evidence jurists and
philosophers about whether the various legal burdens of proof can or should
be theoretically explained as quantified subjective or objective probabilities

as proof by a preponderanceof the evidence.. .. However, in some instances, the burdenof

proof requiresa partyto producea substantiallygreaterdegree of belief in the mind of the trier
of fact concerningthe existence of the fact-a burdenusuallydescribedby statingthatthe party
must produceclear and convincing proof . .. or, with respectto the prosecutionin a criminal
case, proof beyond a reasonabledoubt ....
CAL. EVID. CODE ? 500 cmt. (West 1995). Probabilistsmight suggest that the appropriatelevel of
confidence for the "preponderance" burdenis 51%, for the "clearand convincing"burdensomethinglike
75%, and for the "beyond a reasonable doubt" burden something like 90%. Thus, probabilityas the
standardof epistemic appraisalyields a very clear orderingsystem for these differentlevels of confidence.
This is true whetherthe probabilisticstandardis consideredto be objective, as in assessmentsof relative
frequencyover repeatedtrials, or is consideredto be the subjectiverationalassignmentof betting odds.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1654 The Yale Law Journal [Vol. 107: 1535

(thick commitment) or whether juries should be instructed to think of them in

those terms (thin commitment). Anti-probabilists claim that nonprobabilistic
standards of epistemic appraisal and attendant levels of confidence best
represent the juridical factfinding process, and one can indeed find many
statements by judges that seem to cohere significantly better with a
nonprobabilistic standard. One court, for example, stated that the
"[p]reponderance of the evidence is sufficient if it inclines an impartial and
reasonable mind to one side rather than the other."389We must do a bit of
interpretive work to discern the standardof epistemic appraisal that underwrites
this court's assertion. Even so, it seems fairly clear that the standard of
appraisal presupposed by the court's statement (something like "assessment by
an impartial and reasonable mind") is not intended to be probabilistic, and that
the level of confidence (perhaps something like "inclination to believe, all
things considered"), is not intended to be quantified in probabilistic terms
either.
As suggested above, probabilistic standardsof epistemic appraisal (whether
subjective or objective) can yield different levels of confidence ordered on the
scale set by the standard-for example, 51% for the "preponderance"burden
of proof, 75% for the "clear and convincing" burden, 90% for the "beyond a
reasonable doubt" burden. Nonprobabilistic statements of epistemic appraisal
are also capable of yielding different levels of confidence to reflect different
burdens of proof. For example, different levels of confidence are possible
under the standard of epistemic appraisal "assessment by an impartial and
reasonable mind." Suppose the burden of proof that the law requires is not
"preponderance of the evidence," but instead, "clear and convincing evidence"
(the strongest burden in the civil setting). Under the "impartial and reasonable
mind" standard, a court (or legislature) might require that the jury not simply
be "inclined" to believe the truth of the proposition at issue, but something
stronger, perhaps "firmly inclined." And were the case a criminal one, for
which the "beyond a reasonable doubt" burden of proof was mandated by a
legislature and court, the level of confidence measured by the "impartial and
reasonable mind" standard of epistemic appraisal might be something like
"conviction to a moral certainty."
Anti-probabilists offer nonprobabilistic standards of epistemic appraisal
(and attendant levels of confidence) even when interpreting the not infrequent
assertions made by various kinds of jurist that expressly use such terms as
'probable' and 'likely' in explicating the different burdens of proof. One finds

389. Moss-Am., Inc. v. Fair Employment Practices Comm'n, 317 N.E.2d 343, 351 (Ill. App. Ct. 1974).
A similar standard of epistemic appraisal and level of confidence is used in Livanovitch v. Livanovitch, 131
A. 799 (Vt. 1926), in which an appellate court held that the trial court properly instructed the jury regarding
the requirements of the preponderance of evidence test with this charge: "If . .. you are more inclined to
believe from the evidence that he did so deliver the bonds to the defendant, even though your belief is only
the slightest degree greater than that he did not, your verdict should be for the plaintiff." Id. at 800.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1655

an example of this kind of probabilistic assertion in a venerable and influential

evidence treatise that declares:

The most acceptable meaning to be given to the expression, proof by

a preponderance, seems to be proof which leads the jury to find that
the existence of the contested fact is more probable than its
nonexistence. Thus the preponderance of evidence becomes the trier's
belief in the preponderance of probability.390

One finds many similar statements by courts and other official bodies and
commentators.391Anti-probabilists maintain that even these apparently overt
references to probabilistic judgments are best interpreted as using the language
of probability to convey a nonprobabalistic concept. For example, Charles
Nesson (a leading anti-probabilist) argues that probabilistic terminology in
standards of proof are best understood as expressing political-moral ideals, not
mathematizable calculations:

My thesis, broadly stated, is that "probability" as we use the term

in law, particularly in the civil standard of proof, is not a hard-edged
mathematical concept. It is, rather, a concept that incorporates less
rigid ideas of justice and reflects the judicial function of resolving
disputes in the real world, where values shift and knowledge is
uncertain. An outcome is "probable" if it best accomplishes a just and
acceptable resolution of the dispute. Probability, as a legal concept
in the law of proof suggests wisdom, probity, and approbation-not
favorable betting odds.392

Again, one would have to do some rational-reconstructive work to discern the

precise standard of epistemic appraisal and level of confidence recommended
by an anti-probabilist approach of this sort. Indeed, an important project in
"legal epistemology" (beyond the scope of the present work) is to examine the
different analyses courts and commentators offer of the various burdens of
proof, carefully unpack the presupposed standards of epistemic appraisal and
attendant levels of confidence, and assess them from both epistemic and
practical points of view. Notice also that a given jurisdiction-court or
legislature-would probably select one standardof epistemic appraisal for both
civil and criminal trials, but vary the level of confidence required for either
conviction, in the criminal setting, or verdict for the plaintiff, in the civil.

390. MCCORMICK,supra note 386, at 794.

391. See, e.g., CAL. EVID. CODE. ? 500 cmt. Overt reference to probabilistic standardsis also
frequently encounteredin the judgment of relevance, a necessary condition (under most U.S. rules of
evidence) of admissibility.
392. Charles Nesson, Agent Orange Meets the Blue Bus: Factfindingat the Frontierof Knowledge,
66 B.U. L. REv. 521, 521 (1986) (emphasisadded).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1656 The Yale Law Journal [Vol. 107: 1535

One final note about step (1). The consequent of (1) speaks of judgments
true for given practical purposes. This phrase is best understood not as the
claim that facts are actually different from different points of view, but rather
as the claim that different points of view can require different levels of
confidence and different standards of epistemic appraisal to conclude that a
putative fact is a fact for the aims at hand. Premise (1) focuses on the
deferring practical reasoner's point of view.

3. Commenton Steps (2) to (10)

Step (2) is a premise that unpacks the judgment that practical reasoners
implicitly make when they adjudge something to be true for their purposes.
Step (3) presents the principle that allows the practical reasoner to move, in the
inferential chain, from the conclusion that he ought to infer a judgment
(because it is the result of an expert's scientific process that satisfies L, the
level of confidence of the appropriate standard of epistemic appraisal that the
practical reasoner has specified) to the judgment itself. The judgment thus
inferred is in the right form to serve as the minor premise of the primary
practical syllogism. Because the inference of j from the conclusion that the
practical reasoner ought to inferj is so closely linked to the idea of truth for
given practical purposes, we might call this the "principle of practical purpose
rationality."
In step (4), the practical reasoner must assess whether the judgments that
the expert offers in testimony as the product of his scientific process, namely
P, satisfy the level of confidence that is referred to in step (1), namely L.
Again, the simple example is of a jury whose chosen standard of epistemic
appraisal is subjective probability, and whose chosen level of confidence is
greater than 50%. This practical reasoner looks to the expert to specify the
level of probabilistic confidence the evidence warrants. If that level is 40%,
then for that jury L is not satisfied. For any number greater than 50%, L would
be satisfied. Step (4) reflects a practical reasoner's judgment that L is satisfied.
The practical reasoner looks to the expert for statements that produce the
inference of step (5). In those statements, the expert tells the practical reasoner
that the expert's scientific process P-already adjudged in Step (1) to meet the
requirements of L-actually produces a given judgment. For example, the
expert might testify: "The method I used to relate back the data from the
defendant's blood alcohol level at the time of testing to his level at the time
of arrest shows that there was a 67% chance that he was driving with a 0.10%
level."393
Step (6) is true by (1), (4), and (5). Step (7) is true by (2) and (6). Step (8)
is a new premise. It reflects the fact that, in the setting of practical epistemic

393. For a brief discussion of a case in which an expert offered this testimony,see supra note 376.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1657

deference I am modeling, the practical reasoner has been presented with a

given practical decisionmaking task (as a judge or jury is presented with the
practical task of determining how applicable laws apply to the conduct of a
party to a civil or criminal proceeding). To have such a task at hand (and no
other overriding or inconsistent task) is to "have those practical purposes."
Note that the clause 'and has no other inconsistent practical purposes' allows
for the defeasibility of practical judgments, a reasoning process that can itself
be further modeled.394
Step (9) is thus true by steps (7) and (8). Finally, step (10) is true by (3)
(the "principle of practical-purpose rationality") and (9). Note that judgment
j is the conclusion of the "secondary" argument that begins with step (1). Step
(10) is the minor premise of the primary syllogism for which the secondary
argument does its work. The conjunction of (1) through (9) as antecedent with
(10) as consequent might be thought of as a lemma to the primary syllogism,
or as a "practical theorem." According to the claim about practical reasoning
that I am labeling "minor premise practical priority," every minor premise in
every practical syllogism in which the practical reasoner lacks requisite expert
theoretical competence, is actually the conclusion of an argument. The chief
major premise of this argument is the practical premise in which the practical
reasoner sets both a standard of epistemic appraisal and a related level of
confidence. In the abstract formulation above, this practical premise was step
(1).

D. Oversimplification in the Foregoing Model of Practical Epistemic

Deference

The foregoing model is too simple to capture adequately that reasoning

process of practical epistemic deference. The chief oversimplification is its
failure to take into account the fact that a nonexpert practical reasoner is often
actually faced with not just one expert, but with several who present competing
testimony about the same factual issue. Thus a critical part of the reasoning
process involved in practical epistemic deference is a decision about which of
the competing experts to believe. This is a decision over and above the
decisions, marked in the model above, about which expert's scientific
procedure meets the standard of epistemic appraisal and attendant level of
confidence set by the practical reasoner.
It is in the crucial reasoning about which among competing experts to
believe that the real problems for epistemic deference appear. At least two
significant problems-both identified in Section IV.C-face the deferring
nonexpert, problems that must be resolved in a way that is sufficiently
epistemically warrantedfor practical (legal) purposes, i.e., from the practical

394. See Brewer,supra note 15, at 1017-21.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1658 The Yale Law Journal [Vol. 107: 1535

(legal) point of view: the problem of selection of the particular expert and the
expert scientific discipline and the related problem of competition among
experts.
The problems of selection and competition are formidable, and one or
more of them is faced by every nonexpert reasoner who solicits scientific
information from experts. No model of the process of practical epistemic
deference can be adequate if it does not reflect the rational decision procedures
nonexperts use to resolve these problems. In this way, the model offered above
is deficient in its explanation. To remedy the deficiency, I must return to the
inference form of abduction and explain its crucial role in practical epistemic
deference.

E. Breaking the Dialectical Impasse and Completing the Model:

Abduction in Legal Reasoning

Abduction is a pattern ubiquitous in legal reasoning. I have discussed

elsewhere at some length its basic structure, its logical limits (as an invalid
form of inference), and its concomitant pragmatic strengths (as a rationally
disciplinable method of discovery).395 I have also shown that it plays a
critical role in the process of analogical reasoning.396I will not repeat my
analysis of analogical abduction here, but instead offer a label for it with only
the briefest summary, to help distinguish it from two other types of abductive
inference. I refer to it as analogy-warranting rule abduction. My task now is
to call attention to two other types of, and roles for, abduction in legal
reasoning, both of which are necessary elements in the epistemic task that
faces one or another of the nonexpert practical reasoners (judge, judge plus
jury, even lawyer) in practical epistemic deference.
The basic pattern of abductive inference, shared by all three of the types
I will identify, is the argument form:

Therefore,

where E)is some explanandum, cDsome explanatory hypothesis (that itself has
a conditional logical structure), and '4 -> e' the proposition that e would
follow from or be explained by 4 if indeed 4 were true or otherwise
adequately warranted. The assertion of 4 in the conclusion of an abduction
marks the fact that the rational abducing reasoner has settled tentatively on 4

395. See id. at 945-49, 978-82, 1021-26.

396. See id.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1659

as the proper explanatory hypothesis.397 Although the pattern of abduction

is invalid, there are ways of "settling" on 4 that are pragmatically valuable.
The aware reasoner understands that the conclusion of a valid deductive
argument is held with certainty, and also understands that the conclusion of an
inductive argument is held only with varying degrees of probability, always
less than 100% (the more cogent the argument, the higher the degree of
probability). Such a reasoner also realizes that 4 is asserted in an abductive
conclusion not as a certain truth, but instead as a tentatively held hypothesis
that is sufficiently likely to be the proper explanation of e that it is worth the
effort of confirming or disconfirming it. She further recognizes that every
abductive conclusion needs some kind of confirmation.
In legal reasoning, there are different types of explananda and explanatory
hypotheses. According to the type of each that appears in the particularcontext
of legal reasoning, the type of abductive inference will itself differ. There are
three types: First, in analogy-warranting rule abduction, the reasoner seeks to
abduce a rule that specifies the relevant similarities between a set of example
cases (sources for the analogy) and some target case about which he has a
question. The rule can specify necessary conditions, sufficient conditions, or
necessary and sufficient conditions (but in legal argument usually one or the
other). What he seeks to explain here is the properly similar (or different)
treatment of the target case or cases as indicated by the relevant features of the
source case or cases.398Second, when the "abduced" explanatory hypothesis
explains an empirical fact, the legal reasoner engages in factual hypothesis
abduction, or, as I shall refer to it, fact abduction (an abbreviation only, since
it is not the fact itself, but a hypothesis that explains the fact that is abduced).
This is the effort to discover a factual explanatory hypothesis, 1D,for some fact

397. Paul Thagardargues that abductionis betterunderstoodas an inferenceto the best explanation
of 9 by cDthan as an inference that 9 follows deductively from (D. See Paul R. Thagard,The Unity of
Peirce's Theoryof Hypothesis, 13 TRANSACriONS PEIRCE SOC'Y 112, 116-17 (1977). Thagard'sargument
is persuasive.Althoughin the text I sometimes treatabductiveinferencein the standardPeirceanmanner,
i.e., relying on a premise '( -> 9,' I do think the relationof ( and 9 is that the ( advancedis, in the
judgment of the abductivereasoner,the best availableexplanationof 9.
398. The following sums up in greaterschematic detail the two types of analogical rule abduction:
Wherey,, Y2,Y3, . .y.n are "target"items;x,, x2, x3, . . . x, are "source"items; F,, F2, F3, . . . F"are shared
characteristics;and H,, H2, H3, . . . Hn are inferredcharacteristics,the task of analogy-warrantingrule
abductionis to "discover,"for some items xi and yi, and some characteristicsFi and Hi, a rule of the form
"if thereis anything[in the jurisdiction]thathas F and also has H, then everything[in thejurisdiction]that
has F also has H"-when a premise stating"x is an F andx is an H" is partof the argument,this abduced
rule in turnentails a rule of the form "All F are H." Disanalogy-warranting rule abductioninvolves much
the same structure, with the following changes. In addition to target, source, shared, and inferred
characteristics,G,, G2, G3, . . . Gnare unsharedcharacteristicsthat the source item(s) has, but the target
item(s) does not. The task of disanalogy-warranting rule abductionis to "discover,"for some items xi and
yi, and some characteristicsFi, Gi, and Hi, a rule of the form "by itself, the presenceof F in an item is not
a sufficient condition of H, but the presence of F and not-G arejointly sufficient conditionsfor H." Note
that reasonersabduce disanalogy-warranting rules when there is some prima facie reason to believe that
the compareditems xi andyi do, by virtueof sharingcharacteristicsFi, also sharethe inferredcharacteristics
Hi, even though the reasoner's ultimate conclusion is that only x, and not also y, has the inferred
characteristic.See generally Brewer,supra note 15.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1660 The Yale Law Journal [Vol. 107: 1535

or event, G. Of course, as always with abductive inferences, once abduced, G

must be confirmed or disconfirmed. Fact abduction is the type of abduction
familiar in philosophy of science, and it is what came to Peirce's mind when
he himself "abduced" abduction. It also plays a powerfully important role in
legal reasoning and, more specifically, in the process of practical epistemic
deference I have been exploring.
A third type of abduction plays a role in legal reasoning generally as well
as in practical epistemic deference more specifically. We may call it legal rule
abduction. In this type of abduction, the reasoner is once again faced with a
set of facts that call for explanation. And, as in factual hypothesis abduction,
the explananda are considered to be facts that occurred in the world. But the
type of explanation sought here is not empirical. It is not an explanation from
a scientific point of view (an explanation that proceeds by selecting cognitive
aims and methods that both serve those aims and produce scientific
judgments). Instead, legal rule abduction aims to explain facts with special
regard to the aims, methods, and judgments specific to legal reasoning as a
distinct rational enterprise.399It is a type of reasoning every lawyer must do
when a client walks in the door complaining of something that happened to
him (e.g., someone failed to exercise due care on the sidewalk in front of his
house). In the American federal system, such reasoning is also required of a
judge before she can dismiss a complaint for "failure to state a claim on which
relief can be granted."4wUnited States rules of pleading allow a very vague
summary of the facts that are the basis of complaint. They do not require the
early characterization of those facts in legal terms-that is, a well-pleaded
complaint need not present the litigant's own legal explanation of the facts.
When it does not present such an explanation, the judge must seek to abduce
a legal rule that might give the complainant a sustainable cause of action if the
facts alleged are later proven. Legal rule abduction has this particular structure:

P
if T then (If P then Q)
Therefore,
T,

where T is a plausible theory of the case as well as a "valid" legal rule of the
jurisdiction, one that links the facts a plaintiff claims he can prove to a remedy
he desires in that jurisdiction; P is the set of facts the plaintiff thinks he can
prove (or the judge thinks the plaintiff might be able to prove); Q is the
remedy the plaintiff desires (or the judge thinks the plaintiff might desire). The
reasoner's task in legal rule abduction is to "abduce" the legal rule (the legal
theory, i.e., I) that makes "legal sense" of the facts of the case. As with all

399. See supra Sections II.B-C.

400. FED.R. Civ. P. 12(b)(6).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1661

abductive inferences, the rule abduced must be confirmed or disconfirmed;

what it means to "confirm" a legal rule (or legal theory) that has been abduced
is to establish that the legal rule (or theory) is a "valid" rule in the
jurisdiction.40'
All three types of abduction can take place in a complex pattern of legal
reasoning in a single case. There are especially close holistic connections
between legal rule abduction and factual hypothesis abduction. One such
connection arises from the strategic need of the lawyer to abduce (and then
argue to the court) only those legal theories (if T then (If P then Q)) whose
required factual predicate (P) he thinks he can prove if and when necessary.
There is another connection between fact abduction and legal rule abduction
that is of direct concern to my project here. On the division of labor made in
the American federal system, the judge decides whether a given complaint
brought by a plaintiff states a cause of action upon which relief can be granted.
In deciding that a complaint does state such a cause, the judge perforce sets
the parameters for the decision about what kinds of scientific testimony will
be rationally pertinent to the case. Now, the factfinder has the task of abducing
the factual hypothesis that would best explain the plaintiff's contention. In
Brown,402 the factfinder (ultimately, the Court itself) abduced the factual
explanation that state-mandated segregation caused psychological harm.403
Next, having already abduced and confirmed the legal rule reflected in the

401. An example will help. In Searightv. New Jersey,412 F. Supp.413 (D.N.J. 1976), a troubledman
went to court, representinghimself and complainingthat the state of New Jersey more than a dozen years
earlierinjectedhim with a radiumelectricbeam as the resultof which he heardvoices talkingto him inside
of his brain.He came into federalcourt,althoughit was doubtfulwhetherthe federalcourthadjurisdiction
over his case since therewas neitherclear federalquestionjurisdictionnor diversityjurisdiction.Moreover,
the state statuteof limitationson his tortclaim had run.Yet thereis a rule that requiresa judge to consider
all reasonable,plausible grounds for a cause of action before dismissing the claim. In effect, this rule
requiresthe judge to try to abduce the best legal rule thatmight be adducedto explain the facts of the case.
With a sneering tone unbefittinga judge, this federaljudge did indeed abduce several possible federal
questiontheoriesof recoveryfor the troubledplaintiff.The judge firstconsideredwhetherfederal statutory
civil rights law might afford him a remedy (no, because the plaintiff must first bring the action in state
court when it does not "riseto a constitutionallevel," and this case, in the judge's view, it did not so rise).
Here is the sneering part:The judge then concluded that the facts taken as true (the judge is requiredto
assume that the facts pleaded are true at this stage of litigation) show at most "unlicensed radio
transmission" by someone, id. at 414, but this was within the sole jurisdiction of the Federal
CommunicationsCommission,and so the case did, after all, have to be dismissed.
The judge's best effort to abduce a legal rule that would "explain"this litigant'scase from a legal
point of view in a way that would allow the case to proceed failed to produce such an explanation.In
schematicform, the abductionin Searightlooks like this: Plaintiffclaims P, that while he was in custody,
defendant injected him with a radiumbeam, that this was unlawful under some relevant overall legal
theory, T, and consequentlegal rule, If P then Q (that is, the plaintiffclaims that a valid legal rule in the
jurisdiction is one of the form T, -> (P -> Q), and that the judge's abductivetask is to try to discover
whetherthere really is such a rule in the jurisdiction);and that he was therebyentitled to legal remedy Q,
i.e., money damages of $12 million. The judge's abductive task is to discover, if possible, some legal
theory and correspondingrule that might yield a remedy.Two "abduced"candidatesfor testing:(1) (T. -4
(P -4 Q?)); (2) (T2 -+ (P -+ Q?)); where T, = violation of civil rights law; and T2= violation of FCC
regulations.The judge then works to confirmor disconfirmP-and in this case disconfirmseach candidate
T, and T2.
402. Brown v. Board of Educ., 347 U.S. 483 (1954).
403. See supra Section I.B.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1662 The Yale Law Journal [Vol. 107: 1535

Equal Protection Clause and its attendant doctrines (often the abduction of
legal rules requires the court to resolve vagueness, and often vagueness is
resolved by the use of exemplary reasoning),4'04the Court granted the
plaintiff the remedy sought.405 In the underlying Daubert case,40 the
factfinder (ultimately, the Ninth Circuit Court of Appeals) abduced a factual
explanatory hypothesis regarding the existence of the plaintiff's birth defects.
Having abduced the factual explanation that the defendant's drug caused that
injury, but then having disconfirmed that explanation, the court applied the tort
law (products liability law) it had already discerned and confirmed in legal rule
abduction to give the defendant the victory.407In both Brown and the Ninth
Circuit's final Daubert opinion, the factfinder turned to scientific experts to
abduce explanations of the facts alleged in the complaint.408
Recall that I came back to the topic of abduction having reached an
impasse in thinking about how the nonexpert practical reasoner can go about
resolving selection, competition, and underdetermination problems in a
legitimate manner, when it seems that such resolution can be achieved only by
a reasoner who has the degree of epistemic competence that a nonexpert, ex
hypothesi, does not have.409How can abduction help us here? In the main,
it will help by revealing to a greater level of analytical detail than the
oversimplified model offered in Section VI.C just what it is that the nonexpert
reasoner must do to perform selection tasks in the face of competition by the
experts.
Abduction guides practical epistemic deference in several interconnected
ways. The judge must first abduce and confirm that there is a legal theory
valid in the jurisdiction that could give the complainant the remedy desired if
the complainant can prove the facts asserted. The reasoner (judge or judge plus
jury) must then hear the proof of the facts and decide whether the facts are as
alleged. This task, as argued in the previous section, requires that the reasoner
both abduce and then confirm or disconfirm an explanation of the facts from
a legal point of view. Sometimes this leads the practical reasoner to consult a
scientific expert, depending on the nature of the theory of the case the lawyers
and the judge have already abduced using legal rule abduction.

404. In the context of legal reasoning,analogy-warrantingrule abductionis a special type of legal rule
abduction.
405. See Brown, 347 U.S. at 495.
406. Daubertv. MerrellDow Pharms.,Inc., 727 F. Supp. 570 (S.D. Cal. 1989).
407. See supra Section I.B.
408. The process by which this scientific informationis absorbedinto the practicalreasoner'soverall
reasoning, which I have called "minor premise practical priority,"is outlined supra Section VI.B. It
involves the practicalreasoner(court or legislature,sometimes dividing labor with a jury) adoptingsome
standardof epistemic appraisal,some attendantlevel of confidence, and then relying on it in soliciting
expert information from scientific experts whose disciplines the practical reasoner has judged to be
rationallypertinentto the case at hand.
409. See supra Section V.D.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1663

When expert testimony is available, the nonexpert must make several

closely related judgments, each of which is reflected in the major premise
of the nested syllogism, but some of which are not captured in the
oversimplified model presented in Section VI.C. These include: (1) a
judgment about the proper standard of epistemic appraisal (usually set by
court or legislature for the factfinder); (2) a judgment about the proper
attendant level of confidence (also usually set by court or legislature under
the rubric of the burden of proof); (3) an abductive judgment about the
factual hypothesis that best explains the case; (4) a judgment about whether
there is one or more expert scientific disciplines that are rationally pertinent
to the case and thus ought to be consulted in aid of that abductive inference;
and (5) when the answer to (4) is yes, a selecting judgment about which
purveyor of an expert discipline that passes the test of (4) ought to be
deferred to, even when there is actual or implied competition among the
experts whose testimony has been admitted.
I contend that practical reasoners (sometimes operating as a group with a
coordinated division of epistemic labor-as between judges and juries, for
example) must make this complex network of judgments in every act of
practical epistemic deference. Although these judgments can be modeled in
various ways, I propose the following model, which marks a first step toward
bringing the simple model offered above to a level of complexity that can
adequately capture the reasoning process:

(1) If an expert's scientific process (call it P) (a) produces a

judgment (call it J) that rises to the level of confidence of the
appropriate standard of epistemic appraisal specified by the practical
reasoner (call the specified level 4), and (b) those judgments are
rationally pertinent to the factual question at issue, and (c) P offers
the best explanation of the factual question among the live
competitors, then those judgments are true for given practical
purposes (call them R).

Most of the reasons for the additional components of (1) are offered in the
discussion just above it. One point worth mentioning here is that the judgment
in (1)(c) reflects an understanding of abductive inference as an inference
among live competitor explanations. This is perhaps implicit in the original
Peircean formulation, but it becomes more explicit-and rightly so-in recent
treatments.410The abductive model offered in the artificial intelligence work
done by John Josephson, for example, expressly treats abductive inference as
a choice among competing explanations:

410. Thagard'sargumentis much to the same effect. See Thagard,supra note 397.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1664 The Yale Law Journal [Vol. 107: 1535

(1) D is a collection of data (facts, observations, givens).

(2) H explains D (would, if true, explain D).
(3) No other hypothesis can explain D as well as H does.
Therefore,
(4) H is probably true.4'

Even more to the point for my purposes, Josephson's account is sensitive

to the role of standards of epistemic appraisal and levels of confidence in
abductive inference generally. Discussing what it takes to confirm or
disconfirm an abduced explanation, he offers the following criteria:

(1) How decisively H surpasses the alternatives; (2) How good H is

by itself, independently of any consideration of the alternatives (we
should be cautious about accepting a hypothesis, even if it is clearly
the best one we have, if it is not sufficiently plausible in itself); (3)
Judgments of the reliability of the data; (4) How much confidence
there is that all plausible explanationshave been considered (how
thorough was the search for alternative explanations); (5) Pragmatic
considerations,includingthe costs of being wrongand the benefitsof
being right;(6) How strongthe need is to come to a conclusionat all,
especially seeking further evidence before deciding.412

Josephson's criteria (4) and (6) cohere especially well with observations
and arguments I have made throughout this Article. Criterion (4) reflects the
importance within the theory of abductive inference of recognizing what I have
referred to as implied competition. It is especially useful to keep implied
competition in mind as we seek to explain and understand practical epistemic
deference, since the practical reasoner cannot make a cogent judgment simply
by confining her judgment to the experts that the lawyers happen to offer.
Though this is a tricky matter, I am strongly inclined to the view that rational
abductive inference requires that the reasoner make some judgment about other
expert (or nonexpert) disciplines that might offer superior competing
explanatory accounts of the data.
Criterion (5) calls attention to the importance of setting a level of
confidence that is appropriate to the type of case presented.413Josephson's
model is also friendly to my contention that nonprobabilistic standards of

411. See John R. Josephson, Conceptual Analysis of Abduction, in ABDUCTIVE INFERENCE:

COMPUTATION,
PHILOSOPHY, 5 (John R. Josephsonet al. eds., 1994).
TECHNOLOGY
412. Id. at 14 (emphasis added). Josephson appears to treat criteria (1) to (4) as going to "[tihe
judgment of likelihood associated with an abductiveconclusion,"and criteria (6) and (7) as a separate
consideration(one "[b]eyond the judgment of ... likelihood")regardingthe reasoner's"willingness to
accept the conclusion."Id. I do not see these as separatetypes of consideration,so I have included them
on one list in the text above.
Criterion5 is very close to the point Rudnermakes in his ultimatelyunsuccessfulargumentthat the
scientist as such must make moral decisions. See supra Section II.C.
413. See supra Section II.C (discussing Rudner).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] ScientificExpertTestimony 1665

epistemic appraisalare often chosen by the practicaldecisionmakerin step

.414

Before showing the value of this additionally"complexified"model of

practicalepistemic deference,let me presentthe whole model in accordwith
the additionaljudgmentsreflectedin step (1):

(1) If an expert's scientific process (call it P) (a) produces a

judgment (call it j) that rises to the level of confidence of the
appropriatestandardof epistemic appraisalspecifiedby the practical
reasoner (call the specified level b), and (b) those judgments are
rationallypertinentto the factual question at issue, and (c) P offers
the best explanation of the factual question among the live
competitors, then those judgments are true for given practical
purposes (call them R).
(2) If j is true for R, then if a practicalreasonerhas R (and has
no other inconsistentpracticalpurposes),then that reasonerought to
inferJ.
(3) If a practicalreasonerought to inferj, thenj.
(4) P producesjudgmentsthatsatisfy D and thosejudgmentsare
rationallypertinentto the factual question at issue and P offers the
best explanationof the factual question among the live competitors.
(5) j is the result of P.
(6) j is true for R.
(7) If a practicalreasonerhas R (and has no other inconsistent
practicalpurposes),then that reasonerought to inferj.
(8) The practicalreasonerhas (singularpracticalpurpose)r (and
has no other inconsistentpracticalpurposes).
(9) The practicalreasonerought to inferj.
Therefore,
(10)j.

Step (4) suggests that the nonexpertreasonercannot avoid taking epistemic

responsibilityfor judgmentsregarding:(1) the rationalpertinenceof an expert
method; (2) the required standard of epistemic appraisal and level of
confidence; and, perhapseven more importantly,(3) the conclusion that one
of several competingexpertmethodsproduces the best available explanation
of the fact at issue. In light of this more complex model of deference to
experts,two strongclaims lead me to doubtthata nonexpertcan performthese

414. He writes:
It has been suggested that we should use mathematicalprobabilitiesto help us choose
among explanatoryhypotheses .... If suitable knowledge of probabilitiesis available, the
mathematical theory of probabilitiescan, in principle, guide our abductive evaluation of
explanatoryhypotheses to determinewhich is best. However, in practice it seems that rough
qualitativeconfidence levels on the hypothesesare enough to supportabductions,which then
produce rough qualitative confidence levels for their conclusions .... [Flor the most part
numericalconfidenceestimatesare unavailableand unnecessaryfor reasoning.People are good
abductivereasonerswithoutclose estimates of confidence.
Josephson,supra note 411, at 26-27.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1666 The Yale Law Journal [Vol. 107: 1535

tasks in a manner that is legitimate from the practical point of view of the
legal system.
To articulate that doubt, three steps remain for this Article. First, I
conclude my earlier discussion of the problems of selection and competition
by bringing this more complex model to bear on it. Second, I show that, on
balance, we have strong reason to doubt that the judgments made in steps (1)
and (4) can be made in a manner that is not arbitrary from an epistemic point
of view. Finally, in Part VII, I show that, from a legal point of view, an
epistemically arbitraryjudgment in the process of practical epistemic deference
is not legitimate.
We have already wrestled with conflicting intuitions about a nonexpert's
ability to acquire KJB from experts in the face of problems of selection and
competition.415 It seemed on the one hand that credentials must be sufficient
to enable KJB to arise from practical epistemic deference, but it also seemed
that the use of credentials was subject to forceful challenge insofar as such use
occasioned regress, question begging, and underdetermination. How can
abduction help illuminate this problem?
First, keep in mind that we do not expect that the nonexpert can acquire
KJB from the expert based on second guessing the substantive judgment of the
expert.416Also, keep in mind that we are assuming sincerity and good faith
on the part of the competing experts.417Thus the nonexpert is attempting to
discover which discipline to select and which expert (among competing
experts, actual or implied) to believe on the basis of the credentials (including
the reputation) of sincere experts. What is it that could get this judgment about
credentials, and about credentials as an index to warranted expert beliefs about
the world, off the ground? The nonexpert must consult his own experience,
memory, and prior judgments; where else can he begin? "[W]hat every juror
ought to do in arriving at a verdict," as one court aptly explains, is

use all his experience, his knowledge of human nature, his knowledge
of human events, past and present, his knowledge of the motives
which influence and control human action, and test the evidence in the
case according to such knowledge and render his verdict accordingly.
A juror who did not do this would be remiss in his duty. The trial
judge in considering the verdict must do the same, or fail in the
discharge of that function which the law has laid upon him.418

This means that what the practical reasoner (judge or judge plus jury) must
bring to the table regarding credentials is some explanatory judgment about
the relation between credentials and scientific expertise. In modem society,

415. See supra Section IV.C.

416. See supra SubsectionV.C.1.
417. See supra SubsectionV.C.4.
418. State v. Hammond,604 A.2d 793, 795 (Conn. 1992).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1667

almost all of us have experience with those who appear to us to be well-

credentialed and expert in some discipline. We encounter teachers with
credentials (more credentials as we climb the academic food chain), doctors
with credentials, other professionals with credentials, television experts with
credentials. Moreover, we have some reasonably confirming and disconfirming
experiences with the success of these credentialed experts at resolving
particular problems. We acquire the habit of extrapolating, by abductive-cum-
inductive inference, from our success with one expert's credentials to an
expectation of success with another expert who has the same credentials.
The nonexpert practical reasoner brings all this (implicit) reasoning about
credentials with her to the judge's chamber or jury box. Such reasoning can
indeed serve us well in everyday life, and it can account for the widely held
view (in accounts like those of Kenny, Hardwig, Putnam, and Coady) that
credentials are an adequate device by which a nonexpert can acquire KJB from
an expert. It may even be that these considerations about the role of credentials
in everyday life, and the extrapolation from that role into the domain of
practical legal reasoning, are sufficient for the purposes of the legal
system.419 But there are at least three reasons to doubt seriously that these
considerations are adequate and, though it is a close question, I am inclined to
think they are not.
First, what is good enough for oneself in one's own life in the way of
believing credentials and their experts is often not sufficient for deciding the
fate of someone else who is being hauled about in civil or criminal process
with life, liberty, property, or reputation at stake. (Explaining this suggestion
is part of the burden of the next section.) Second, the kinds of expert
judgments that are rationally pertinent to legal decisions are not infrequently
at the cutting edge of scientific theory (for example, DNA fingerprinting,
causation of harm by cigarettes or other consumer products, or causation of
psychological harm), or at the cutting edge of particular scientific methods
(which themselves cannot be kept sharply distinct from theories). Even
applications of scientific methods that involve widely confirmed and accepted
methods often involve complex mathematical formulae, including probabilistic
and statistical judgments that are notoriously counterintuitive. As suggested
above, to use credentials as an index of expertise, the nonexpert must
extrapolate from prior experience with and judgments about credentialed
experts. But even this generalization itself can be accomplished only with the
help of an abductive inference, namely an inference to a hypothesis that

419. I should acknowledge that my approachto the question of the reasoner'sexperience with the
success of credentialsand extrapolationtherefromdraws its energy from a moderateempiricistapproach.
Detailed discussionof why I am attractedto thatapproachis beyondthe scope of the presentwork. I would
begin by defending an empiricistaccount with the kinds of modificationscogently suggested by Fricker.
See Fricker,supra note 239.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1668 The Yale Law Journal [Vol. 107: 1535

explains the pattern of successes and failures with prior credentialed experts
as purveyors of KJB about the world.
I assume that this abductive inference can be folded into judgment (1)(c)
in the complex model, but it could be exploded and analyzed into its own
distinct set of steps. Also, I should explain that every inductive generalization
relies on abductive inference to generate the categories to be tested and
confirmed or disconfirmed by observation. The reason for this partakes of an
observation also relied on in my discussion of analogy elsewhere:420Every
item has an indefinite number of features, from which some enormously
narrowing selection must be made to get to the point of applying some method
like John Stuart Mill's methods of confirmation or disconfirmation. A white
swan is also a feathered creature, a biped, a fleshy creature, a carbon-based
creature, a noncarnivorous creature, a creature with at least one limb, a
creature without gills, a creature that weighs less than the earth, and a creature
that travels more slowly than the speed of light. She who would like to
confirm some truth about swans, such as that they are all white, or that most
of them are white, must select from this welter of characteristics and place
those characteristics selected into an explanatory pattern that is ready-made for
observation, verification, and falsification. That selection is abductive
inference.42' The upshot of this is that no nonexpert can inductively
generalize from experience with credentialed experts without relying on a
logically prior step of abduction.
It is difficult to see how the nonexpert has a large enough base of
experiences with experts to make good plausible abductive-cum-inductive
inferences about the epistemic "warrantingness" of credentials. Even a
successful experience with a doctor may not be best explained by the fact that
he went to a good medical school; maybe a midwife or a witch doctor would
have done as well because the ailment was psychosomatic and responded to
a placebo effect.422Indeed, in a skeptical mood, one might well undermine,
or at least raise serious doubts about, virtually every conclusion a nonexpert
makes about the KJB-producing capacity of credentials. But even in a less
skeptical mood, we should acknowledge that a nonexpert's base of extra-
judicial, pre-cameral experience with credentialed experts is likely to be small,
and some significant portion of it is likely to be shaky.
Finally, even apart from the adequacy of the base of data on which the
nonexpert would rely in abducing a hypothesis about credentials, there is the
far more daunting problem of confirmation or disconfirmation of whatever
hypothesis is abduced. All abductions require confirmation or disconfirmation,

420. See Brewer,supra note 15, at 932-33.

421. Josephsonalso offers an argumentto show that"it is possible to treatevery good (i.e., reasonable,
valid) inductive generalizationas an instanceof abduction."Josephson,supra note 411, at 19.
422. See the interestingdiscussionin JEROME D. FRANK, PERSUASION AND HEALING: A COMPARATIVE
STUDY OF PSYCHOTHERAPY 136-64 (2d ed. 1973).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific ExpertTestimony 1669

but the nonexpert'slack of epistemic competencethreatensto deprive her of

precisely the kind of understandingshe would need to be able to confirmor
disconfirm a hypothesis about credentials and their capacity accurately to
identify which expertsare capableof producingKJB and which are not. Such
a hypothesiswould includejudgmentsabouthow importantit is to the process
of KJB formationfor the expertto have a degree, which kind of degree, how
importantreputationis distinct from other credentials, and in what way
reputationis important.
Somehow, with this last consideration,it seems I have come full circle
(hoping not to have argued that way). It is difficult to see how a
nonepistemicallycompetentpracticalreasonercould make a cogent abductive
inference about the KJB-signalingcapacity of credentials, and confirm or
disconfirm it, without having epistemic competence in the credentialized
disciplines. In the absence of some reason to believe that the reasonercan
make a cogent abductionof this sort, and in the absence of some reason to
believe thatthe nonexperteven possesses a sufficientbase of dataon the basis
of which to abduce a plausiblehypothesis(aboutwhich credentialsare KJB-
signaling) in the first place, the judgmentby the nonexpertin relianceon that
abductionseems epistemicallyarbitrary.
I have just arguedthat in at least some cases in which competingexpert
testimony is presented,a nonexpert'sjudgment in reliance on an abduction
about credentials will be arbitrary.The question naturallyarises: In what
percentageof cases? A fully skepticalview on this questionwould be thatsuch
judgmentsare inevitablyarbitrary;thatis, given these epistemicconstraintson
the nonexpert,no judgment about scientific evidence that itself relies on an
abductive inference about the KJB-signalingpower of credentials can be
nonarbitrary.Although I am somewhat inclined toward this more skeptical
view, I should emphasize that I do not need to go that far to establish that
thereis considerableepistemicarbitrariness in nonexperts'judgmentsthatrely
on abductionsabout credentials.I call this the "moderatelystrong"skeptical
view, in contrast to the immoderatelyskeptical route of asserting that all
judgments by nonexperts about credentials are epistemically arbitrary.
According to more moderateskepticism,judgmentsabout experts' scientific
propositionswill be arbitraryat least when the following conditionis satisfied:
Whenever the criterion of credentials underdetermineswhat scientific
propositionis endorsed-that is, when two roughlyequally well-credentialed
experts (to the eyes of the nonexpert)endorsecompetingpropositions,either
about scientific theory, about scientific method, or about the applicationof
scientific theory and methodto a particularcase.423Let us refer to this as the
"underdetermination condition."Although it would be difficult to estimate

423. See the discussion of Laudan'sreticularaccount of scientific inquiry,which I track here. See
supra Section II.B.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1670 The Yale Law Journal [Vol. 107: 1535

accurately the exact percentage of cases in which the underdetermination

condition is in fact satisfied in the litigation setting, or in other settings, it is
clearly a very significant percentage. Indeed, we should recognize that relevant
kinds of disagreement between competing experts (i.e., those disagreements in
which roughly equally well-credentialed experts endorse competing
propositions) occur at all three levels that comprise an expert "point of
view"-namely, the level of theory, the level of method, and the level of
application. Recognizing that fact about experts' disagreement allows us to see
that the underdetermination condition is satisfied in a far greater percentage of
scientific testimony cases than we would notice were we to focus only on
experts' disagreements about theory or method. For example, a very large
percentage of cases in which scientific testimony is presented deal with
"accident reconstruction" in tort suits.424In these cases, both the theory and
basic methods of science are usually not in dispute; what is in dispute is which
expert's application of accepted theories and methods (mostly physics and
biology) provides the best explanation of a particular incident. There is
competing testimony by roughly equally well-credentialed experts in a large
percentage of these cases.
Thus, according to the moderately strong skeptical position I am here
endorsing, there are serious epistemic defects in the process by which a
nonexpert extrapolates from prior (pre-judicial, if not prejudicial) experience
with the success and failure of credentials as signals of KJB production. These
epistemic defects are such that, when faced with competing, sincere, and
roughly equally well-credentialed experts (i.e., when the underdetermination
condition is satisfied), a nonexpert will on average do no better in selecting

424. An article in a leading litigator's handbook describes "accident reconstruction"and expert

testimony in this way:
Wheneverone is confrontedwith difficult liability issues in a substantialmotor vehicle injury
case, a dissection of how the accident happened,is essential. This is particularlysignificant
when the victim is one who is involved in a multi-vehicle crash situation. Under those
circumstances,a firm graspof the dynamicsof speed, roadsurface,metal damage,skid marks,
gouge marks,debris, scuffing of roadway,and other factors, are paramountin reconstructing
the events that led to the accident in question.The burdenof reconstruction(which falls upon
the party assertingthe claim or defense) can be met in but one of two methods of proof: The
first option is to produce facts through witnesses and documents that will create sufficient
inferences, thereby drawing the jury to the desired conclusion. The other methodology is to
producea professionalwitness called an "accidentreconstructionspecialist."Such an expert is
one who is usually a licensed professionalengineerwith a specialty in the myriadof dynamics
involved in a complex motor vehicle accident. Essentially, such a witness, once qualified,
provides testimony that is usually identicalwith the ultimateor thresholdissue of the case. In
essence, such a witness furnishes the accusatoryopinion that incriminatesone or the other
operatorinvolved in the accident. While the argumentcan be raised that the ultimateopinion
reachedby such an expert is a usurpationof the jury function, it remains,notwithstanding,as
opinion evidence, subjectto the usual instructionby the Courtas to its acceptanceor rejection
by the jury. Moreover, strong and direct accident reconstructiontestimony is extremely
persuasiveand oftentimes representsthe differencebetween winning and losing.
LeonardL. Finz, Accident Reconstruction:A VerySpecial Art, in LITIGATION177, 179-80 (PLI Litig. &
Admin. PracticeCourse HandbookSeries No. H4-5012, 1987).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1671

which scientific expert to believe than one would by tossing a coin. That is,
when a nonexpert is faced with competing testimony by roughly equally well-
credentialed experts, and that nonexpert does manage to acquire a true belief
about the scientific matter at hand, we have no reason to believe that the true
belief was arrived at other than by accident.
There is enormous complexity among competing theories of knowledge,
but for all the very significant differences among them, in this they seem to
converge: A method of acquiring belief that produces only accidentally true
beliefs is epistemically arbitrary and incapable of producing KJB. It is hard to
improve upon Bernard Williams's succinct statement of this view:

[I]f we are speaking in general about knowledge... [w]hat is

necessary-and what represents the undoubted fact that knowledge
differs from mere true belief-is that one or more of a class of
conditions should obtain, which relate the fact that A has this belief
to the fact that the belief is true: conditions which can best be
summarised by the formula that, given the truth of p, it is no accident
that A believes p rather than not-p. This formula is vague and
over-generous, b:t it gets us, I think, on the right line ... .425

Time to sum up. I have argued that reliance on the credentials of (even)
sincere competing experts requires the nonexpert to abduce and to confirm or
disconfirm a hypothesis about the KJB-producing capacity of credentials as
signals, but that the nonexpert's lack of epistemic competence will prevent him
from executing the first step in the chain of practical epistemic deference in a
nonarbitrary manner, at least when the "underdetermination condition" is
satisfied, that is, when roughly equally well-credentialed experts offer
testimony that competes at the level of theory, method, or application.
Moreover, competition by such experts occurs in a fairly large percentage of
cases of scientific testimony-especially since there is large room for
disagreement about application even when the basic theory and method are not
in dispute among the relevant experts. The defect in the first step will cascade
through the rest of the steps (2) to (10), rendering the whole chain arbitrary,
from an epistemic point of view, when the "underdeterminationcondition" is
met. The final question for us is: What if anything has the legal system to say
about practical epistemic deference that is epistemically arbitrary?

425. B.A.O. Williams, Knowledge and Reasons, in PROBLEMS IN THE THEORY OF KNOWLEDGE 1, 5
(G.H. von Wright ed., 1972). In the omitted parts of this passage, Williams firmly commits himself to an
externalist approach to knowledge.
[N]ot only is it not necessary that the knower be able to support or ground his true belief by
reference to other propositions, but it is not necessary that he be in any special state with regard
to this belief at all, at least at the level of what he can consciously rehearse.
Id. Elsewhere, the paper flirts with a causal account. I would subscribe to neither epistemological view, but
I think virtually all theories would reject the merely accidentally true belief as either justified belief or
knowledge.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1672 The Yale Law Journal [Vol. 107: 1535

VII. INTELLECTUALDUE PROCESS

One value of the closely related ideas of the axiological structureof an

intellectual enterpriseand the point of view426is that they provide us with
a clear way to examine the interactionof the reasoningof practicalreasoners
and theoreticalreasonerswhen the formersolicit expertinformationfrom the
latter.These concepts facilitatethe separationof the practical aims, methods,
andjudgmentsof a legal reasonerfrom expert theoreticalaims, methods,and
judgments.It is especially importantfor us to have some such analyticaltool
because, on the surface, the distinctivepoints of view of the expert scientist
and the judge or jury can seem hardto keep separatewhen the judge or jury
seeks in the course of legal decisionmakingto make or at least endorse
theoreticaljudgments(in cases like Brownand Daubert).

A. EpistemicNonarbitrarinessas a Practical Constrainton

LegitimateEpistemicDeference

My goal for the present section is to explicate some of the overall

normativeaims in the "practicalpoint of view" from which a legal system
ought to evaluatethe transferof scientific informationfrom scientificexperts
to nonexpertjudges andjurors.The normativeaims thatare such an important
element in the practicalpoint of view of a legal system are articulatedand
relied upon in many decisions by courts dealing in one way or anotherwith
the rationalityof legal decisionmakingin cases to which complex scientific
informationis rationallypertinent.I discuss a few such cases to help explicate
what those normativeaims are.
The centralidea animatingthese practicalnormsis thatcertainrule-of-law
values requireepistemicnonarbitrarinessin factfindingreasoning,as in other
types of reasoning.Thus,if the nonexpertcannotacquirescientificbeliefs from
competing experts in a way that is nonarbitrary, from an epistemic point of
view, those beliefs will thereforenot be legitimate from the practical legal
point of view. That is, accordingto this practicalrule-of-lawnorm,at least in
cases in which life, liberty,or propertyis at stake, epistemicnonarbitrariness
in the process of "finding" scientifically discerned facts is a necessary
condition of the practical legitimacy of a decision that relies on that
factfinding.
One finds respect for and recognitionof this norm in both philosophical
and legal materials.JohnRawls, for example,has long maintainedthat among
those "guidelinesintendedto preservethe integrityof the judicial process"are
the requirementsthat courts undertakeconscientiously

426. See supra Sections II.B-C.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998) ScientificExpertTestimony 1673

to determinewhetheran infractionhas takenplace .... Thus a legal

system must. .. contain rules of evidence that guarantee rational
proceduresof inquiry.While thereare variationsin these procedures,
the rule of law requiressomeform of due process: that is, a process
reasonablydesignedto ascertainthe truth,in ways consistentwith the
other ends of the legal system, as to whethera violation has taken
place and underwhat circumstances.427

Rawls's point appliesno less to the integrityof thejudicialprocessin civil

cases thanit does to the integrityof thatprocessin criminalcases. A reflection
of this same basic rule-of-law value, articulatedin the setting of a civil
antitrust action, is found in a prominentfederal appellate case. In In re
Japanese ElectronicProductsAntitrustLitigation,428several U.S. electronics
manufacturers brought an antitrust action against several Japanese
manufacturers.When the U.S. plaintiffs made a motion for jury trial, the
defendantJapanesecompaniescountered,claimingamongotherthingsthatthe
economic and technical issues were too conceptuallycomplex for the jury to
understand,even with the help of expert testimony.The case was on appeal
from a federal districtcourt on the sole question whetherthe plaintiffshad a
right to a jury trial even when the issues and evidence involved would be
acutely complex.429In effect, the court consideredwhetherthere was (what
has come to be referredto as) a "complexity exception" to the Seventh
Amendment right to jury trial under the U.S. Constitution.430 The district
judge had held that there was no such "complexityexception."431 In what
appearsto be an unprecedenteddecision in the federal courts, the appellate
court overturnedthe districtcourtjudgment.The appellatecourt agreed with
the defendant,vacatedthe districtcourt'spretrialorder,and held that, despite
the normativeforce of the Seventh Amendmentright to jury trial, the Fifth
Amendment due process right to have a rational and fair adjudication
outweighedthe SeventhAmendmentright.The courtconcludedthat the Fifth
Amendmentnarrowsthe scope of the Seventh Amendmentby means of a
complexity exception.
In its cogently articulatedopinion, the appellatecourt specified the kind
of complexitythat might trumpthe right to jury trial as follows:

A suit is too complex for a jury when circumstancesrenderthe jury

unable to decide in a propermanner.The law presumesthat a jury
will find facts and reach a verdict by rational means. It does not
contemplatescientific precisionbut does contemplatea resolutionof

427. JOHNRAwLs, A THEORYOF JUSTICE238-39 (1971) (emphasisadded).

428. 631 F.2d 1069 (3d Cir. 1980).
429. See id. at 1071.
430. See id. at 1079-80.
431. See id. at 1073 (discussing Zenith Radio Corp. v. MatsushitaElec. Indus. Corp., 478 F. Supp.
889 (E.D. Pa. 1979)).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1674 The Yale Law Journal [Vol. 107: 1535

each issue on the basis of a fair and reasonableassessment of the

evidence and a fair and reasonableapplicationof the relevantlegal
rules. A suit might be excessively complex as a result of any set of
circumstances which singly or in combination render a jury unable to
decide in the foregoing rational manner Examples of such
circumstances are an exceptionally long trial period and conceptually
difficultfactual issues.432

The court concludes that "due process precludes trial by jury when a jury
is unable to perform this task with a reasonable understanding of the evidence
and the legal rules."433 The court went on to elucidate the connection
between this right to rational comprehension by the legal decisionmaker and
rule-of-law values like predictability and notice:

The due process objections to jury trial of a complex case implicate

values of fundamental importance. If judicial decisions are not based
on factual determinations bearing some reliable degree of accuracy,
legal remedies will not be applied consistently with the purposes of
the laws. There is a danger that jury verdicts will be erratic and
completely unpredictable, which would be inconsistent with
evenhandedjustice. Finally, unless the jury can understand the
evidence and the legal rules sufficientlyto rest its decision on them,
the objectiveof most rules of evidenceand procedurein promotinga
fair trial will be lost entirely. We believe that when a jury is unable
to perform its decisionmaking task with a reasonable understanding
of the evidence and legal rules, it undermines the ability of a district
court to render basic justice.434

The court also addressed the question of what values the legal system might
injure in cases to which the complexity exception applied by choosing not to
allow the kind of community input that the constitutional jury trial right was
designed to secure. The central values often mentioned in connection with the
jury trial right-values powerful enough, in the district court's judgment, to
lead it to reject the idea of a complexity exception-are the jury's function as
a check on judicial power, and the jury's ability to modify and conform the
law to, and suffuse the law with, community values (so-called jury equity),
thereby lending the law a communitarian legitimacy it might not otherwise
have. Assessing these countervailing values reflected in the Seventh
Amendment, the court of appeals delivered something of a coup de grace
debater's point:

432. Id. at 1079 (emphasisadded) (citationsomitted).

433. Id. at 1084.
434. Id. (emphasisadded).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1675

In the context of a lawsuit of the complexity that we have posited,

however, these features [of the jury system] do not produce real
benefits of substantial value. The function of "jury equity" may be
legitimate when the jury actually modifies the law to conform to
community values. However, when the jury is unable to determine the
normal application of the law to the facts of a case and reaches a
verdict on the basis of nothing more than its own determination of
community wisdom and values, its operation is indistinguishable from
arbitrary and unprincipled decisionmaking. Similarly, the
"line-drawing" function is difficult to justify when the jury cannot
understandthe evidence or legal rules relevantto the issue of where
to draw a line.
... A jury unable to understand the evidence and legal rules is
hardly a reliable and effective check on judicial power. Our liberties
are more secure when judicial decisionmakers proceed rationally,
consistently with the law, and on the basis of evidence produced at
trial. If the jury is unable to function in this manner, it has the
capacit2 of becoming itself a tool of arbitrary and erratic judicial
power.

In In re Japanese ElectronicProductsAntitrustLitigation,one thus finds

a powerful articulation of the legal system's commitment to practical norms
in the family of rule-of-law values that are specifically addressed to the
epistemic cogency of juridical factfinding. In various ways, one finds a similar
commitment to these epistemically oriented rule of law values in many other
judicial opinions, state and federal. Thus, one state supreme court declared:

One cogent reason for overturning the verdict of a jury is that the
verdict is based on conclusions that are physically impossible. "[A]
verdict should be set aside '[w]here testimony is thus in conflict with
indisputable physical facts, the facts demonstrate that the testimony is
either intentionally or unintentionally untrue, and leave no real
question of conflict of evidence for the jury concerning which
reasonable minds could reasonably differ.' . . ..
Scientific evidence is relevant to a determination of what is
physically impossible. In Roma v. Thames River Specialties Co., this
court held that the trial judge "would have failed in his duty" if he
had not set aside the verdict when "the laws of mechanics, as testified
to and uncontradicted, tended to prove [the claimant's] story
impossible." In Jump v. Ensign-Bickford Co. the trial court properly
set aside the verdict when expert scientific testimony indicated that it
was physically impossible for a fuse to burn as fast as the claimant
had alleged, and this court could "find in the evidence no reasonable
ground which would have justified the jury in disregarding that
evidence."436

435. Id. at 1085 (emphasisadded) (citationsomitted).

436. State v. Hammond,604 A.2d 793, 795 (Conn. 1992) (citationsomitted).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1676 The Yale Law Journal [Vol. 107: 1535

Similarly, in an opinion perhaps signaling that the U.S. Supreme Court may
be amenable to something like a complexity exception, Justice Souter declared
that, when a case requires the legal decisionmaker to interpret and comprehend
complex technical patents, the decision is to be made by the judge, not the
jury. He reasoned:

In the main, we expect, any credibility determinations will be

subsumed within the necessarily sophisticated analysis of the whole
document, required by the standard construction rule that a term can
be defined only in a way that comports with the instrument as a
whole. Thus, in these cases a jury's capabilities to evaluate
demeanor, to sense the "mainsprings of human conduct," or to reflect
community standards ... are much less significant than a trained
ability to evaluate the testimony in relation to the overall structure of
the patent. The decisionmaker vested with the task of construing the
patent is in the better position to ascertain whether an expert's
proposed definition fully comports with the specification and claims
and so will preserve the patent's internal coherence.437

These cases suggest a strong commitment among leading jurists to the idea
that factfinding, including factfinding regarding matters that are the special
epistemic province of expert scientists, must be conducted in a coherent and
rational manner in order that this epistemic process meet the normative
requirements of a legal system that operates to grant or deprive people of life,
liberty, and property. Reflecting on these and other such statements by judges
and other jurists, we may sense a commitment, immanent in the broad
materials that constitute authoritative law (U.S. law, at least), to what we may
call the practical norm of intellectual due process. In re Japanese Electronic
Products Antitrust Litigation is especially fertile here, with its suggestion that
the epistemic process of comprehension of theoretical complexities is a
mandate of the decidedly practical norm of the Due Process Clause. Though
there are, to be sure, many features of due process that do not specifically
"sound" epistemic, that case reveals that some distinctively practical norms do
have meaningful epistemic consequences. Much philosophical work remains
to be done to explicate this emerging rule-of-law norm. Recognition of
intellectual due process as a practical norm in the family of rule-of-law norms
has only just begun-partly because the startling advances in scientific
methods-coupled with the striking increases in the technological complexity
of society and the laws that arise to govern and guide it, is also a relatively
recent socio-epistemic phenomenon.

437. Markmanv. Westview Instruments,Inc., 116 S. Ct. 1384, 1396 (1996) (citationsomitted).From
an epistemic point of view, Markman'ssolution suffers from the same problemthatafflicts In Re Japanese
ProductsAntitrustLitigation:A technically nonexpertjudge is not in a decisively betterposition than a
technically nonexpertjury.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] ScientificExpertTestimony 1677

A great deal of work remains in explicating the scope and criteria of

"intellectualdue process,"but this much seems clear even now: A reasoning
process that is epistemicallyarbitraryis incapableof producinga legitimate
decision, for such a reasoningprocess is "indistinguishablefrom arbitraryand
unprincipleddecisionmaking."If I am rightthatpracticalepistemic deference
to expertscientistsis doomed,on average,to generatein nonexpertjudges and
jurorsbeliefs that are only accidentallyand arbitrarilytrueat best and thus are
not epistemicallyjustifiedbeliefs, then this process perforcedoes not produce
legally legitimatedecisions.

B. Consequencesfor Doctrinal and InstitutionalDesign:

"Two-Hat"Solutionsand IntellectualDue Process

What is to be done?Detailedanalysisof the consequencesof this analysis

for institutionalanddoctrinalrevisionandtransformation arebeyondthe scope
of my currentproject.A few conclusionsdo emerge,however.If legal systems
are to endorse and aspire to satisfy the intellectualdue process norms (and
other relatedrule-of-lawnorms),they would be well advised to move toward
a "two-hat"model of legal decisionmakingin areasto which scientificresults
are rationallypertinent.On this model, the system seeks to ensurethatone and
the same decisionmakerhas both legal legitimacy (by being duly elected or
appointed by a legitimate elective or appointingauthority)and epistemic
competencewith the basic formaltools of scientificanalysis.A useful heuristic
analogy might be that of a mathematicianor physicist who has practical
decisionmakingauthorityas a voting memberof his department(wearingone
hat), and epistemic competencethat informsthe practicaljudgment(wearing
the other), or similarly,a physicist voting on who should receive a physics
prize for the most importantcontributionto his field. Manyjuristshave already
begun to considerdifferentpaths on this broadavenue of reform,and various
proposals consistent with the "two-hat solution" satisfy it. These include
turning over many decisions currentlymade by private litigation to public
administrativeagencies staffed with trainedscientists,relying on blue ribbon
scientificallytrainedjuries, scientificexpertmagistratejudges, or even special
science courts staffed by scientificallytrainedjudges. Alreadyin the wake of
Daubert's increaseddemandson federaltrial courtjudges, special workshops
on scientific theory and methodhave become availableto trainthem.
A furtherword about Daubert's gatekeepingsolution is in order.Both
Daubert and In re Japanese ElectronicProductsAntitrustLitigationrely on
the underlying assumptionthat a judge is in a decisively better epistemic
position than a jury to assess rationallythe merits of competing scientific
testimony, even when the underdetermination condition is satisfied.438My

438. See supra SubsectionV.C.4.c (describingthe "underdetermination

condition").

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1678 The Yale Law Journal [Vol. 107: 1535

analysis of the legitimacy of epistemic deference gives reason to be skeptical

about that assumption, for the analysis applies no less to a nonexpert judge
than it does to a nonexpert jury. It is for this reason that the distinction
between the threshold question of the admissibility of evidence, on the one
hand, and the question of the weight of the evidence, on the other, is not
particularly important in my analysis. Daubert and (implicitly) In re Japanese
Electronic Products Antitrust Litigation treat this distinction as very important,
for they both assume that a judge is in a significantly better epistemic position
to decide whether proffered scientific evidence is sufficiently reliable to be
admissible in a trial before a nonexpert jury, which could then weigh the
suitably screened evidence. I have argued that we have good reason to doubt
that assumption and, indeed, to be quite skeptical about the idea of solving
problems of selection, competition, and underdeterminationby taking decisions
about expert testimony away from nonexpert juries and giving them to
nonexpert judges.
It is important not to overstate my argument here. Early in the Article, I
noted that epistemic competence is a matter of degree-that not all experts are
equally epistemically competent and not all nonexperts are equally
epistemically incompetent.439 This means that it is certainly conceptually
possible that a trial judge is significantly more epistemically competent than
a jury in assessing the scientific merits of expert scientific testimony, even
when the underdetermination condition is satisfied. That is, it is conceptually
possible that the underlying assumption of Daubert, In re Japanese Electronic
Products Antitrust Litigation, Markman, and many other state and federal
cases, is accurate as to some judges. Nor is this a bare conceptual possibility.
It is not unreasonable to suppose that some judges, who are repeatedly and
predictably faced with proffers of scientific evidence, may find and take the
time and energy required to become decently competent in manipulating the
aims, methods, and results of some of the specific sciences that are likely to
come into their courts. Perhaps some autodidactic judges even become
sufficiently competent to satisfy the demands of intellectual due process. The
plausible possibility that this is true of some judges raises the largely empirical
question about what percentage of judges in state or federal systems are in fact
in this happy-from the point of view of intellectual due process-state.440
Still, though I have not done the kind of empirical work required cogently
to answer that question, the norm of intellectual due process itself places the
burden of empirical proof on those who would maintain that a large enough
percentage of judges are or will in the near future be in that state. That is, the
burden is on the person who claims, along with the Daubert Court, that, by
and large, trial judges already wear, or soon will wear, the required "two hats."

439. See supra Section IV.A.

440. Thanks to Justice Charles Fried for very helpful discussion of this point.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] ScientificExpertTestimony 1679

The burdenis on the person who claims that the requirementsof intellectual
due process can be satisfiedon a large scale by takingthe decisions out of the
hands of nonexpertjuries and leaving them in the hands of judges. Carrying
thatburdenof proofwould of coursecall for a proceduresignificantlydifferent
from that of Daubert itself, for Daubert is still willing to turn over many
ultimatedecisions, even in cases in which the underdetermination conditionis
satisfied, to nonexpertswhom we have no reason to believe are sufficiently
competentin the expertdisciplineto meet the requirementsof intellectualdue
process.
I have spoken of remaining empirical questions and of burdens of
empiricalproof. There are also importantconceptualdetails thatremainto be
worked out for the two-hat solution. What kind of training should the
experts-or expertly trainedjudges-get? One can get a clear sense of the
training requiredto have basic competence in biology, genetics, statistics,
economics, or epidemiology,but how scientificallyspecializedis it feasible to
allow the two-hat-wearinglegal decisionmakerto be? Even the heuristic
analogy of the mathematicianmay breakdown, since that field, like virtually
all fields in the empiricaland demonstrativesciences, is becoming intensely
specialized. Will scientific discipline become so specialized that it ceases to
make sense to talk about general epistemic competence even within a
discipline? And if that problemlooms for singulardisciplines, what hope is
there for resolving problems of extra-disciplinarycompetition in a manner
consistentwith the suasionsof intellectualdue process?Still anotherquestion
is, how much training is enough? To the level of a Ph.D? An M.A.? Are
formal degrees good signals at all? Still another question involves the
democraticlegitimacyof the two-hatsolution.Rule by technocrat-kingshas its
dangers,just as does rule by epistemicallyunrulymobs. Is therea feasible and
meaningfulway in which a responsiblepolity can deliberateand endorsethe
trainingprogramsand institutionalschemes that would implementthe "two-
hat" solution and achieve a reasonabledegree of intellectualdue process?
These are deep anddifficultquestions.But few topics, I hazardto say, will
be more importantto the health of the polity and its citizens than the close
investigation of how the law ought-from legal, moral, and other closely
relatedpracticalpoints of view-to keep up with science.

VIII. CONCLUSION

Though its steps have been long, the argumentpresentedin this Article is
not too difficult to summarize. I have argued that there is a structured
reasoningprocess that a nonexpertjudge or jury must use in an effort to take
accountof scientificexperttestimonyin the courseof reachinga legal decision
aboutliability (in the civil setting)or guilt (in the criminalsetting).When one
attends carefully to the precise steps of the reasoningprocess, one sees that

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1680 The Yale Law Journal [Vol. 107: 1535

there are crucial steps that a nonexpert judge or jury is, in a great many
instances, not capable of performing in an epistemically nonarbitrarymanner.
Specifically, when competing scientific experts are, for all the nonexpert
knows, fairly evenly matched in credentials, reputation, and demeanor, and
when no generally accessible rational criteria (such as self-contradiction by an
expert witness) break the "tie" (i.e., when what I have called the
"underdetermination condition" is satisfied), then a nonexpert is not capable
of choosing among the competing experts in an epistemically nonarbitraryway.
I have also sought to show that epistemic nonarbitrariness is a necessary
condition of legitimacy, as expressed in the norm of intellectual due process.
This norm, an emerging rule-of-law norm, immanent in both decided cases and
various analyses of jurists and philosophers, will be increasingly important as
scientific expert testimony comes to be used in a greater and greater percentage
of cases.441When the conditions of this norm are not satisfied, decisions by
nonexperts, even in light of relevant scientific expert testimony, lack epistemic
legitimacy, and therefore lack the kind of practical-cum-moral legitimacy that
legal systems do and ought to demand.
What is to be done if the relevance of scientific information to legal
decisions continues to grow, while nonexperts are so often incapable of
legitimately incorporating that information into their decisions? Nothing in this
Article suggests that a nonexpert judge cannot become sufficiently
epistemically competent, even without the formal training of a scientist.
Perhaps some judges, by virtue of background or repeat "on the bench"
experience with scientific evidence, will become sufficiently epistemically
competent to render decisions about scientific expert testimony that are
epistemically legitimate and that meet the demands of intellectual due process.
Daubert itself calls upon judges to be more active as "gatekeepers" in
screening out unreliable science. But Daubert's solution to the problem of
legitimately assessing expert scientific testimony seems a poor one. First, given
the press of other judicial business, it seems unlikely that a significant
percentage of judges either already have, or will find the time to acquire, the
kind of scientific competence that legitimate, intellectually "duly processed"
decisionmaking requires. That is, unless judges are routinely and systematically
trained in scientific theories and methods, Daubert does not offer a promising
overall solution to the problem. Also, under Daubert, even when a judge is
sufficiently competent, that competence could yield a duly processed,
legitimate decision only when the judge decides not to admit some proffered
scientific testimony. But in a great many other cases, the judge will admit
competing scientific evidence, and allow the nonexpert, noncompetent jury to

441. See JUDICIALCONFERENCE OF THEU.S., REPORTOF THEFEDERALCOURTSSTUDY COMMITTEE

97 (1990) ("Economic, statistical, technological, and natural and social scientific data are becoming
increasingly important in both routine and complex litigation."), quoted in General Elec. v. Joiner, 118 S.
Ct. 512, 520 (1997) (Breyer, J., concurring).

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions
1998] Scientific Expert Testimony 1681

make the decision-the quintessential circumstance, as this Article has argued,

in which failures of intellectual due process occur.
Nor can the expedient emphasized by Justice Breyer in the Joiner
decision-the Supreme Court's next major treatment of an issue concerning
scientific evidence after Daubert-resolve the problem. Justice Breyer rightly
highlighted the epistemic problem judges have in attempting to fulfill
Daubert's requirements:

This requirement will sometimes ask judges to make subtle and

sophisticated determinations about scientific methodology and its
relation to the conclusions an expert witness seeks to
offer-particularly when a case arises in an area where the science
itself is tentative or uncertain, or where testimony about general risk
levels in human beings or animals is offered to prove individual
causation. Yet... judges are not scientists and do not have the
scientific training that can facilitate the making of such decisions."2

In Justice Breyer's view, given this challenge, judges should more actively
solicit scientific information in order to perform their Daubert-mandated role
of "gatekeepers" vigilant against junk science, for example by using court
appointed experts,443 appointing special masters or specially trained law
clerks, or using pretrial conferences to narrow the scientific issues."4 But,
for reasons suggested above, this solution also fails to meet the needs of
intellectual due process for any judge who is himself not epistemically
competent in scientific methods and theories.445For the judge is not capable
of making an epistemically legitimate decision about which special master, law
clerk, or court-appointed expert to consult.
The only solution (actually, it is a family of solutions) I see requires that
one and the same legal decisionmaker wear two hats, the hat of epistemic
competence and the hat of practical legitimacy. That is, whether it is a
scientifically trained judge or juror or agency administrator, the same person
who has legal authority must also have epistemic competence in relevant
scientific disciplines. In an age in which culture will increasingly take
advantage of the massive intellectual power of science, this is not too high a
price for the legal system to pay to satisfy its own just intellectual aspirations.

442. Id. Joiner held that, even underDaubert, the properstandardof review for decisions about the
admissibility of scientific evidence was "abuseof discretion,"regardlessof whether the districtjudge's
decision was to admit or exclude the evidence, and regardlessof whether that decision was "outcome
determinative."See id. at 515 (majorityopinion).
443. See FED. R. EvID. 706.
444. See Joiner, 118 S. Ct. at 520-21 (Breyer,J., concurring).
445. Like the philosopherAnthonyKenny,Justice Breyeroffers an "extra-cameral" approachto the
problem.I discuss this approachand its problemsabove. See supra notes 286-293 and accompanyingtext.

This content downloaded from 200.16.5.202 on Mon, 26 Jan 2015 18:46:47 PM

All use subject to JSTOR Terms and Conditions