Causation in Tort Law

Richard W. Wright

TABLE OF CONTENTS

I. THE SCOPE AND SIGNIFICANCE OF THE CAUSAL INQUIRY

IN TORT LAW ............................................ 1741
A. The Scope of the CausalInquiry: Distinguishingthe
CausalInquiry from the Tortious-Conduct and
Proximate-CauseInquiries ............................ 1741
1. M alone ........................................... 1742
2. Hart and Honori ................................. 1745
3. Epstein ........................................... 1750
B. The Significance of the CausalInquiry: Linking the
CausalInquiry and the Tortious-ConductInquiry
Through the Tortious-Aspect CausationRequirement.... 1759
1. Green's Duty-Risk Theory ......................... 1761
2. Keeton's "Harm Within the Risk" Theory .......... 1763
3. The Tortious-Aspect CausationRequirement ........ 1766
4. The Power of the Tortious-Aspect Causation
Requirement ...................................... 1771
II. THE NATURE AND CONTENT OF THE CAUSAL INQUIRY ... 1774
A. The But-For Test and Its Limits: Duplicative and
Preemptive Causation ................................. 1775
B. Efforts to Modify the But-For Test ..................... 1777
1. Detailing the Manner of Occurrence ............... 1777
2. Detailing the Injury ............................... 1778
3. Excluding Hypothetical Facts ...................... 1780
4. Aggregating Multiple Potential Causes.............. 1780
C. The Substantial-FactorFormula ....................... 1781
D. Undefined, Directly Observable Causal Contribution .... 1784
E. The NESS (Necessary Element of a Sufficient Set) Test. 1788
1. The Philosophic Basis of the NESS Test ............ 1789
2. The Duplicative-CausationCases ................... 1791
3. The Preemptive-CausationCases ................... 1794
4. Distinguishingthe Damages Issue: The Successive-
Injury and Overwhelming-Force Cases .............. 1798
5. Theft, Nonuse, or Misuse of Defective or Missing
Safety Devices .................................... 1801
F. The FactualNature of the CausalInquiry.............. 1803
1735
1736 CALIFORNIA LAW REVIEW [Vol. 73:1735

1. The Hypothetical-InquiryArgument ................ 1803

2. Malone's Arguments .............................. 1807
III. CAUSATION, PROBABILITY, AND RISK: PROBLEMS ON THE
FRONTIERS OF TORT LIABILITY ........................... 1813
A. The Reduced-Chance and Increased-Risk Cases ........ 1814
B. The Alternative-CausationCases ....................... 1816
C. DistinguishingProbabilityStatements from Causal
Attribution ........................................... 1821
CONCLUSION ................................................... 1826
1985] CA USA TION IN TORT LAW 1737

Causation in Tort Law

Richard W. Wrightt

In all of tort law, there is no concept which has been as pervasive

and yet elusive as the causation requirement, which relieves a defendant
of liability if his tortious conduct was not a cause of the plaintiff's injury.'
Although described by no less an authority than William Prosser as one
of the "simplest and most obvious" problems in determining tort liabil-
ity,2 the causation requirement has resisted all efforts to reduce it to a
useful, comprehensive formula and has been the subject of widely diver-
gent views concerning its nature, content, scope, and significance.
Through the early part of this century, judges applied causal analy-
sis very broadly. They used causal rhetoric to determine not only
whether the defendant's tortious conduct had contributed to an injury
but also whether the defendant should be held legally responsible for the
injury in the light of the other contributing factors. This broad use of
causal language was sharply challenged during the second quarter of this
century by the legal realists and their colleagues. They insisted that the
only causal issue in a tort case is the simple factual issue of actual causa-
tion or causation-in-fact: whether the defendant's conduct actually con-
tributed to the plaintiff's injury. They noted that most of the courts'
allegedly causal analysis had nothing to do with the issue of causation-in-
fact, but rather, under the misleading label of "proximate cause" or
"legal cause," served as a smokescreen for policy judgments on whether
the defendant ought to be held liable for an injury to which his conduct
had clearly contributed. The actual causation question was considered to
be of little help on this issue of ultimate responsibility, since there are
innumerable causes of each injury. Instead, the realists argued that, in

t Associate Professor of Law, Chicago-Kent College of Law, Illinois Institute of

Technology. B.S. 1968, California Institute of Technology; J.D. 1973, Loyola University, Los
Angeles; LL.M. 1976, Harvard University. I would like to thank George Christie, Donald Marshall,
Paul Shupack, Stewart Sterk, Richard Weisberg, and Elliott Wiss for their helpful comments on the
initial draft of this article.
1. E.g., O.W. HOLMES, THE COMMON LAW 64 (M. Howe ed. 1963); W. PROSSER,
HANDBOOK OF THE LAW OF TORTS § 41, at 236-37, 241 (4th ed. 1971). This statement of the
causation requirement is not meant to exclude injunctive relief against threatened injury. See id. § 1,
at 2. The statement also is ambiguous with respect to whether the causation requirement applies to
the defendant's conduct as a whole or to the tortious aspect of the conduct. This ambiguity is
addressed infra in Part I, Section B.
2. W. PROSSER, supra note 1, § 41, at 237; accord L. GREEN, RATIONALE OF PROXIMATE
CAUSE 4, 132-33, 135-36 (1927); 2 F. HARPER & F. JAMES, THE LAW OF TORTS § 20.2, at 1110-21
& n.4 (1956 & Supp. 1968). Harper and James's chapter 20 is a reprint of James & Perry, Legal
Cause, 60 YALE L.J. 761 (1951).
1738 CALIFORNIA LAW REVIEW [Vol. 73:1735

order to reach the socially advantageous, expedient, or just result in each

case, the policy issues should be openly addressed under the more appro-
priate headings of "duty," "interests protected," "damages," and so
forth. 3
Although the realists were unable to eliminate the use of proximate-
cause language, they were very successful in gaining academic and judi-
cial acceptance of their view that causation-in-fact is the only truly causal
issue, and that proximate-cause determinations are noncausal policy
judgments on the appropriate limits of liability for actually caused
harm. 4
However, courts and legal scholars were unable to formulate a com-
prehensive, policy-neutral account of actual causation.5 These failures
eventually led to the view that the actual causation issue itself is perme-
ated by policy considerations. This view was expressed most forcefully in
an influential article published in 1956 by Wex Malone. He asserted that
the actual causation requirement is just one more doctrinal concept that
is manipulated by judges to further the relevant socially preferred policy
in each case.6 His view has been widely accepted,7 even in limited form
for the most difficult cases by some of the staunchest defenders of actual

3. See L. GREEN, JUDGE AND JURY 29-37, 191, 195-96, 222-25, 230-31, 242-43 (1930); L.
GREEN, supranote 2; W. PROSSER, supra note 1, §§ 41, 42, at 236-37, 239, 244-50 (4th ed. 1971); id.
§§ 45, 46, at 311-21 (1st ed. 1941); see F. HARPER, A TREATISE ON THE LAW OF TORTS § 110, at
257-58 (1933); Edgerton, Legal Cause (pt. 2), 72 U. PA. L. REV. 343, 373 (1924); Gregory,
Proximate Cause in Negligence-A Retreat from "Rationalization," 6 U. CHI. L. REV. 36 (1938);
Morris, On the Teaching of Legal Cause, 39 COLUM. L. REV. 1087 (1939).
4. See, e.g., G. CHRISTIE, CASES AND MATERIALS ON THE LAW OF TORTS 245-46, 264
(1983); R. EPSTEIN, C. GREGORY & H. KALVEN, CASES AND MATERIALS ON TORTS 272-73, 309
(4th ed. 1984) [hereinafter cited as EPSTEIN, TORTS]; J. FLEMING, THE LAW OF TORTS 170-71, 179-
80 (6th ed. 1983); M. FRANKLIN & R. RABIN, CASES AND MATERIALS ON TORT LAW AND
ALTERNATIVES 269, 302 (3d ed. 1983); 2 F. HARPER & F. JAMES, supra note 2, §§ 20.1, 20.2, 20.4;
C. MORRIS & C.R. MORRIS, MORRIS ON TORTS 154-201 (2d ed. 1980); W. PROSSER, supra note 1,
§§ 41, 42, at 236-37, 239, 244-50.
5. See infra text accompanying notes 166-225.
6. Malone, Ruminations on Cause-in-Fact,9 STAN, L. REV. 60 (1956).
7. See, eg., Cole, Windfall and Probability: A Study of "Cause" in Negligence Law (pts. I &
2), 52 CALIF. L. REV. 459, 764 (1964); Delgado, Beyond Sindell: Relaxation of Cause-in-FactRules
for Indeterminate Plaintiffs, 70 CALIF. L. REV. 881, 891-92 (1982); Pedrick, Causation, the "Who
Done It"Issue, andArno Becht, 1978 WASH. U.L.Q. 645; Pound, Causation, 67 YALE L.J. 1 (1957);
Robinson, Multiple Causationin Tort Law: Reflections on the DES Cases, 68 VA. L. REV. 713, 713-
14 (1982); Rosenberg, The CausalConnection in Mass Exposure Cases: A "PublicLaw" Vision of the
Tort System, 97 HARV. L. REV. 849, 855 n.27, 863-64 (1984); Strachan, The Scope andApplication of
the "But For" CausalTest, 33 MOD. L. REV. 386, 389-95 (1970); Weinrib, A Step Forwardin Factual
Causation, 38 MOD. L. REv. 518, 529-33 (1975); Zweir, "Cause in Fact" in Tort Law--A
Philosophicaland HistoricalExamination, 31 DE PAUL L. REV. 769 (1982); see 2 F. HARPER & F.
JAMES, supra note 2, § 20.2 & nn.16 & 17 comments at 92-93, 94-95 (Supp. 1968); W. KEETON, D.
DOBBS, R. KEETON & D. OWEN, PROSSER AND KEETON ON THE LAW OF TORTS §§ 41,42, at 264-
65 & n.7, 279-80 (5th ed. 1984) [hereinafter cited as PROSSER & KEETON]; cf. C. MORRIS & C.R.
MORRIS, supra note 4, at 188-90 (asserting that liability is sometimes imposed for policy reasons
although causation is lacking).
1985] CA USA TION IN TORT LAW 1739

causation as a factual inquiry.' It has been adopted enthusiastically by

the economic analysts of tort law, who argue that the socially preferred
policy in every case is the maximization of social wealth. 9
But there has also been another, very different reaction to the views
of the legal realists, which was initiated by H.L.A. Hart and Tony
Honor6's landmark book, Causationin the Law, originally published in
1959.10 As an alternative to the ad hoc social policy analysis of the legal
realists, Hart and Honor6 attempted to rehabilitate the broad use of
causal analysis to address questions of ultimate liability. They argued
that the ordinary, nonlegal use of causal language provides definite, com-
monsense causal principles that can be, should be, and are used by the
courts to frame factual inquiries which encompass not only the narrow
issue of cause-in-fact but also most of the proximate-cause issues. 1
Indeed, they stated, "causing harm constitutes not only the most usual
but the primary type of ground" for imposing liability.12 Subsequent tort
scholars, most notably
3
Richard Epstein, have attempted to elaborate on
this statement.1

8. See A. BECHT & F. MILLER, THE TEST OF FACTUAL CAUSATION IN NEGLIGENCE AND
STRICT LIABILITY CASES 24,44 n.57, 78 n.109, 81-82 & n. 114, 85-87, 99, 104-06 & n. 157, 112 n.173,
118, 120, 124-26, 128-30, 135-37, 221 (1961); J. FLEMING, supra note 4, at 172-73; H.L.A. HART &
T. HONORt, CAUSATION IN THE LAW lxi-lxii, 7-8, 62, 101-02 & n.46, 239-41, 410, 412-16 (2d ed.
1985); Green, The Causal Relation Issue in Negligence Law, 60 MICH. L. REv. 543, 548-49, 553-55
& n.29, 560-61, 568 (1962); see also W. PROSSER, supra note 1, § 41, at 243 & n.53.
9. See, eg., Calabresi, ConcerningCause and the Law of Torts, 43 U. CHI. L. REV. 69, 69 n.1,
86-87, 105-08 (1975); Grady, A New Positive Economic Theory of Negligence, 92 YALE L.J. 799, 799-
800, 804 n.21 (1983); Landes & Posner, Causation in Tort Law: An Economic Approach, 12 J.
LEGAL STUD. 109, 110-11, 134 (1983); Shavell, An Analysis of Causationandthe Scope of Liabilityin
the Law of Torts, 9 J. LEGAL STUD. 463, 464, 502-03 & n.79 (1980). For my critique of the
economic analysts' views on causation in tort law, see Wright, Actual Causation vs. Probabilistic
Linkage: The Bane of Economic Analysis, 14 J. LEGAL STUD. 435 (1985).
10. H.L.A. HART & A. HONORt, CAUSATION IN THE LAW (1st ed. 1959). The recently
published second edition of this book contains a lengthy new preface but otherwise is generally the
same as the first edition. H.L.A. HART & T. HONORIt, supra note 8. Two changes relevant to the
issues addressed in this Article are discussed infra at notes 12 & 31.
11. H.L.A. HART & A. HONORt supra note 10, at 3-5, 23-25, 61-63, 65, 83-105, 123-25, 261-
62, 273-76; H.L.A. HART & T. HONORt, supra note 8, at xxxiv-xxxvi, lii-liii, 3-5, 24-27, 65-67, 69-
70, 88-111, 130-32, 291-92, 304-07. This objective is stated so emphatically and repeatedly
throughout the book that it is difficult to comprehend Izhak Englard's failure to see it and his
criticism of John Borgo, who did see it, for "misunderstanding" Hart and Honor6. Englard, The
System Builders: A CriticalAppraisal of Modern American Tort Theory, 9 J. LEGAL STUD. 27, 57
n.135 (1980) (criticizing Borgo, Causal Paradigmsin Tort Law, 8 J. LEGAL STUD. 419, 421-25 &
n.17 (1979)). Apparently, Englard does not distinguish between the prevailing academic policy-
oriented interpretation of proximate cause, which Hart, Honor6, and Borgo were attacking, and the
commonsense judgments independent of particular legal policies which Hart, Honor6, and Borgo
argue are controlling in actual adjudication.
12. H.L.A. HART & T. HONORP, supra note 8, at 65; see id. at lxxvii-lxxxi, 66-67, 302.
However, in other passages (especially in the second edition of their book), Hart and Honor6 state
that causing harm usually must be coupled with wrongful conduct to create liability. Id. at xxxv,
xiiii-xlvii, lxxv-lxxvii, 131-32, 254-55, 466.
13. A. HARARI, THE PLACE OF NEGLIGENCE IN THE LAW OF TORTS 49-56 (1962); Epstein, A
1740 CALIFORNIA LAW REVIEW [V/ol. 73:1735

In this Article, I present a systematic critique of these various

accounts of the causation requirement in tort law, 4 and, building on that
critique, I construct a comprehensive new account which hopefully will
be more satisfactory than any of the earlier accounts.
This Article is divided into three parts. Part I discusses the manner
in which the causal inquiry fits into tort liability analysis. Part II
dicusses the nature and content of the causal inquiry itself. Part III dis-
cusses some current problems and controversies dealing with the rela-
tionships among causation, probability, and risk.15
More specifically, I argue in Part I that much of the disagreement
and confusion in the debates about causation stems from failures to dis-
tinguish the causal inquiry from other elements of the courts' liability
analysis, or to pay sufficient attention to the links that connect the differ-
ent elements. As a result of these failures, most writers substantially
overrate or underrate the role that the causal inquiry plays in tort liabil-
ity analysis. Thus, some writers attribute to the causal inquiry issues that
are properly associated with the tortious-conduct or proximate-cause
inquiries. Conversely, other writers trivialize the causal inquiry by link-
ing it to the actor's conduct as a whole, rather than focusing it-as the
courts do--on the tortious aspect of the actor's conduct, and seek to fill
the resulting void with inadequate duty or "harm within the risk" analy-
ses. When the causal inquiry is properly limited and focused, it is seen to
play a powerful but not conclusive role in determining tort liability.
Part II begins with a brief discussion of the traditional "but for"
(necessary condition) test of causation and its inability to account for the
overdetermined-causation cases-cases in which two or more factors
each would have been sufficient to produce the injury, so that none of
them was a necessary condition for the injury. The various tests that
have been proposed as refinements of, supplements to, or substitutes for
the but-for test are then examined. All these tests, with the exception of
the "NESS" (Necessary Element of a Sufficient Set) test, are shown to be
inadequate. The NESS test, which incorporates the traditional Humean
philosophic account of the meaning of causation, as modified by John
Stuart Mill, was first suggested by Hart and Honore. Their brief exposi-
tion of this test, however, was submerged and weakened by their attempt
to treat the tortious-conduct and proximate-cause inquiries as part of the
causal inquiry. In Part II the NESS test is more fully elaborated. It is

Theory of StrictLiability, 2 J. LEGAL STUD. 151, 163, 168-69 (1973). Compare Borgo, supranote 11,
at 425 & n.17, 431 & n.31, 432-40, 452-55 & nn. 55 & 59 (selecting responsible causes) with H.L.A.
HART & A. HONORt, supra note 10, at 26-47, 58-76 (same).
14. For my critique of the accounts of the economic analysts, see Wright, supra note 9.
15. Throughout this Article, the arguments presented will be descriptive rather than
normative. The normative issues will be addressed in a separate article now in progress.
1985] CAUSATION IN TORT LAW

then used to resolve the problematic causation cases that have resisted
solution under all the alternative tests. In the final section of Part II, the
steps involved in applying the NESS test are analyzed to demonstrate
that, contrary to the currently popular view among legal writers, the
causal inquiry is a factual inquiry not dependent on policy
considerations.
In Part III I address the difficult issues raised by the "increased
risk," "reduced chance," and "alternative liability" cases-cases in
which it can be proven that the defendant tortiously exposed the plaintiff
to an increased risk of injury but not that the defendant actually contrib-
uted to the subsequent injury. It is commonly argued that the causation
requirement is properly ignored or relaxed in these cases, or that the
requirement should be satisfied by a probabilistic "increased risk" con-
cept of "causation." I contend that these cases are more plausibly and
fruitfully viewed as cases that recognize a new type of injury-risk expo-
sure-in certain narrowly circumscribed situations. Causation of this
risk-exposure injury is easily established using the basic concept of causa-
tion embodied in the NESS test. Finally, I argue that the probabilistic
increased-risk concept of causation must be rejected, because it lacks the
attributive element that distinguishes causal explanations from mere
probability statements. This attributive element, which has always been
essential for tort liability, explains the courts' refusal to admit pure or
"naked" statistical evidence as proof of causation or identification.

I
THE SCOPE AND SIGNIFICANCE OF THE CAUSAL INQUIRY
IN TORT LAW

A. The Scope of the CausalInquiry: Distinguishingthe Causal

Inquiry from the Tortious-Conductand Proximate-Cause
Inquiries
A number of legal scholars have equated causation of injury with
legal responsibility for the injury." Relying on this assumed equiva-
lence, some of these writers argue that because the determination of
responsibility is policy-dependent, the causal inquiry also must be policy-
dependent. Conversely, others argue that because the causal inquiry is
policy-free, the responsibility determination also must be policy-free.
In this Section it will be demonstrated that both of these competing
arguments are built on a false premise. Causation is not equivalent to
responsibility. Rather, before a decision is made on legal responsibility,

16. See supra notes 12 & 13 and accompanying text; see also Green, supra note 8, at 545, 562;
Malone, supra note 6, at 62-67; cf RESTATEMENT (SECOND) OF TORTS § 431 comment a (1965)
(substantial factor test of causation includes notion of responsibility).
1742 CALIFORNIA LAW REVIEW [Vol. 73:1735

the causal inquiry must be combined with the tortious-conduct inquiry-

"Was the defendant's conduct tortious?"-and the proximate-cause
inquiry-"Is there an applicable policy or principle which absolves the
defendant from liability, even though his tortious conduct was a cause of
the injury?"
As we shall see, the writers who equate causation with responsibility
erroneously attribute to the causal inquiry issues that are properly associ-
ated with the tortious-conduct and proximate-cause inquiries. Three dif-
ferent examples will be discussed. The first is drawn from Wex Malone's
influential article on causation in tort law. In this article, Malone con-
tends that the causation requirement is just one more policy-dependent
doctrine that is manipulated by the judge or jury to reach a desired
result. The second and third examples are Hart and Honore's and
Richard Epstein's respective theories of tort liability, which attempt to
avoid ad hoe, policy-dependent liability determinations by basing liability
on expanded notions of causation.

1. Malone
Malone's article contains a series of arguments that are meant to
establish the policy-dependent nature of the causal inquiry. Only one of
them is pertinent here.17
This argument focuses on the policy considerations that underlie
any decision to identify one of several contributing factors as "the cause"
of an injury. Malone notes that, depending on their respective back-
grounds and motivations, different persons may identify different con-
tributing factors as the cause of an injury. In one of his examples, an
irresponsible youth drives too fast down a road recently covered with
loose gravel, and a stone is thrown by a wheel of the car into the face of a
pedestrian. Malone observes that the neighbors may say that the parents'
letting their irresponsible son drive was the cause of the harm, while a
road engineer may say that the cause was improper road construction,
and a physics teacher may say that the impact of the wheel's momentum
on the rock was the cause. Thus, Malone states, causal determinations
are evaluative and purposive. 8
In Malone's other example, an elderly worker with a long-standing
heart ailment dies of heart failure while engaged in some trivial task for
his employer. A medical expert, says Malone, is likely to testify that the
trivial task was not a cause of the death, since he will view it as an
instance of a commonplace recurrent event with which medical science is
helpless to deal. But a judge, considering the compensatory purposes of

17. Malone's other arguments will be discussed later. See infra text accompanying notes 240-
42 & 304-30.
18. Malone, supra note 6, at 62.
1985] CAUSATION IN TORT LAW

the workers' compensation statute, might decide, or allow a jury to

decide, that the trivial exertion was a work-related cause of the elderly
employee's death. 19 Malone concludes:
"[S]imple" causation is not merely an abstract issue of fact. . . [T]he
resolution of the cause problem depends largely upon the purpose for
which cause is to be used. . . . It is through the process of selecting
what is to be regarded as a cause for the purpose of resolving a legal
dispute that considerations of policy exert their influence in deciding an
issue of cause-in-fact. 20
This argument confuses several distinct parts of the liability analy-
sis. It is generally agreed that the purpose and context of the particular
inquiry determine which of several contributing factors will be selected
as "the cause."2 1 But many writers, including Malone, do not recognize
that the influence of purpose or policy extends only to the noncausal
parts of this selection process: the focus on only one or a few of the
many potential causes, and the limitation of responsibility to only certain
types of consequences.
Thus, in the elderly worker case, policy considerations are quite
important in deciding which causes and consequences the workers' com-
pensation statute was meant to cover. At a minimum, such statutes
require that the employee's work or working conditions have contributed
to (been a cause of) the injury. That is, only work-related activities and
conditions are relevant as potential causes. This is a policy decision. The
trivial exertion was a work-related exertion. But did it contribute to the
failure of the employee's heart? That is the causal question. The causal
issue need not be addressed if it is decided that an injury will not be
treated as work-related, even if the injury occurred while at work,
when-as seems to be the case here-the risk of such injury was not
increased by being at work and, moreover, the risk was personal to the
employee. This also is a policy issue.2 2
If the court decides that such injuries are work-related, the causal
issue must be addressed. The judge should clarify the causal issue for the
medical expert by explaining that the medical controllability of the risk is
irrelevant. The question is rather whether the exertion in fact contrib-
uted to the employee's death by triggering or accelerating the failure of

19. Id. at 63-64.

20. Id. at 64.
21. E.g., T. BEAUCHAMP & A. ROSENBERG, HUME AND THE PROBLEM OF CAUSATION 261-
62, 282, 284-95 (1981); H.L.A. HART & T. HONORt, supra note 8, at lxxvi, 11-12, 35-38, 62; Borgo,
supra note 11, at 430-32, 439-40; Cohen, Field Theory and JudicialLogic, 59 YALE L.J. 238, 251-59
(1950); Cole, supra note 7, at 459-60, 462-65; Mackie, Causes and Conditions, in CAUSATION AND
CONDITIONALS 21-24 & n.15 (E. Sosa ed. 1975); Williams, Causation in the Law, 1961 CAMBRIDGE
L.J. 62, 63-65, 69.
22. See EPSTEIN, TORTS, supra note 4, at 925-27; 1 A. LARSON, THE LAW OF WORKMEN'S
COMPENSATION § 7.20 (1984). See generally id. §§ 6, 7, 12 & 13.
1744 CALIFORNIA LAW REVIEW [Vol. 73:1735

his heart. Malone's causal-selection argument does not demonstrate that

policy considerations do or should play any role in this causal part of the
analysis. All Malone's argument shows is that policy considerations
determine which causes and consequences will give rise to liability.
The elderly worker example is a workers' compensation case, rather
than a tort case. But the liability analysis in tort law proceeds in the
same manner. First, those potential causes that may give rise to tort
liability are identified. We are not interested in all the possible causes,
but only those that were tortious. This is the tortious-conduct inquiry.
Policy considerations determine whether certain conduct will be treated
as tortious.2 3
The second step, after the identification of tortious conduct which
may have contributed to the injury, is the application of the actual-causa-
tion requirement, which requires that the tortious conduct actually have
contributed to the injury.24 This is the causal inquiry. At this stage it is
irrelevant that there may also be other contributing factors (causes).
In fact, it is never necessary for the court to identify or investigate
all the contributing factors. It usually will consider only a few contribut-
ing factors other than the defendant's tortious conduct: those which
might reduce or eliminate, for reasons of policy or principle, the defend-
ant's legal responsibility for harm that was caused by his tortious con-
duct.2 5 This is the proximate-cause inquiry.
All three steps-the tortious-conduct inquiry, the causal inquiry,
and the proximate-cause inquiry-are involved in determining whether
the defendant's conduct was "the cause" of the injury. The causal
inquiry determines whether the defendant's conduct was a cause of the
injury. The tortious-conduct and proximate-cause inquiries determine
whether the defendant should be held legally responsible as "the" cause
of the injury. That is, the phrase "the cause" is simply an elliptical way
of saying "the responsible cause." Malone asserts that the determinations
involved in establishing that a certain factor not only was a cause but also
should be held responsible as "the" cause often are inseparable in the
conversations and minds of ordinary people.26 The distinction may not
be explicit in conversations, but it is certainly implicit. It is essential to
distinguish each step in the adjudication of legal disputes to ensure that
the causal inquiry will focus only on the legally relevant potential causes
and not be confused with the noncausal issues, as occurred in Malone's
23. The traditional types of tortious conduct recognized by the courts are intentional,
negligent, and ultrahazardous behavior.
24. More precisely, the tortious aspect of the conduct must have contributed to the injury. See
infra text accompanying notes 100-64.
25. J. FLEMING, supra note 4, at 171; Green, supra note 8, at 548, 557-59, 564; Williams, supra
note 21, at 63-65.
26. Malone, supra note 6, at 66-67.
1985] CAUSATION IN TORT LAW 1745

discussion of the elderly-worker example.27

2. Hart and Honori

Like Malone, Hart and Honor6 apply the causal label to a combined
tortious-conduct, cause-in-fact, and proximate-cause inquiry. Unlike
Malone, they argue that the combined inquiry is a factual inquiry based
on commonsense causal principles, rather than a policy-dependent
inquiry based on ad hoe judgments of legal purpose or social expedi-
ency.28 However, their commonsense principles are essentially the negli-
gence and intentional tort prongs of the tortious-conduct inquiry and
have nothing to do with causation. The causal label is not only mislead-
ing, but also makes it difficult for them to account for strict tort liability
and certain types of proximate-cause cases.
Hart and Honor6 divide their causal analysis into two steps. The
first step is the cause-in-fact inquiry, which determines whether the
defendant's conduct actually contributed to the injury. If it did, it was a
condition for the occurrence of the injury. 29 The second step is a confla-
tion of the tortious-conduct and proximate-cause inquiries, in which
Hart and Honor6's commonsense "causal" principles are used to deter-
mine whether the defendant's conduct can be distinguished from the
other contributing factors ("mere conditions") as "the cause" of the
30
injury.
According to Hart and Honor6, the central notion in the common-
sense concept of causation is that the cause is the factor which "makes a
difference" by interfering with, intervening in, or otherwise changing the
normal or reasonably expected course of events. Thus, a contributing
factor is treated as the cause rather than as a mere condition if it was (1)
a voluntary human intervention that was intended to produce the conse-
quence (for example, deliberately breaking a vase) or (2) an abnormal
action, event, or condition in the particular context (for example, a freak

27. A. BECHT & F. MILLER, supra note 8,at 5-7, 12-13; Williams, supra note 21, at 63-65, 69.
In another "Rumination" published fourteen years later, but much less well known (outside
Louisiana), Malone reversed his position and insisted that the causal inquiry is purely factual and
should be sharply distinguished from the independent policy considerations that enter into
determination of the responsible cause. Malone, Ruminations on Dixie Drive It Yourself Versus
American Beverage Company, 30 LA. L. REv. 363, 370-71 (1970). For a valiant effort to reconcile
the two Ruminations, see Note, When Cause-in-Fact Is More Than a Fact: The Malone-Green
Debate on the Role of Policy in DeterminingFactual Causationin Tort Law, 44 LA. L. REV. 1519,
1540-41 & n.89 (1984).
28. See supra note 1I.
29. H.L.A. HART & T. HONORt, supra note 8,at 109-11; see id. at 72. Hart and HonorE use
the "necessary element of a sufficient set" test in this cause-in-fact inquiry. See infra text
accompanying notes 226-83.
30. H.L.A. HART & T. HONORP, supra note 8,at 1-2, 24-25, 33, 72-73, 110-11.
1746 CALIFORNIA LAW REVIEW [Vol. 73:1735

storm or driving at an excessive speed).3" If there are two or more con-

tributing factors which satisfy one of these criteria, the last one to occur
is treated as the cause. When searching for the cause of an injury, we do
not trace back any further once we come across a deliberate intervention
or an independent abnormal condition.32
It soon becomes clear that these criteria are neither policy-neutral
nor causal. As applied to human conduct in a tort case, the two criteria
simply call for inquiries into the intentional or negligent character of the
conduct. This is, in effect, a restricted form of the tortious-conduct
inquiry.
Thus, a human intervention is considered to be voluntary only if it
was a free, deliberate, and informed act or omission, intended to bring
about what in fact happened in the manner in which it happened. An
intervention is not voluntary if the consequence was not intended, or
even if it was intended if the intervention occurred as a result of mistake,
ignorance, coercion, or the pressure of moral or legal obligation, or if the
choice to intervene was the lesser of two evils.3 3 More generally, Hart
and Honor6 admit that their voluntary-intervention criterion calls for
31. Id. at 2, 29, 33-35, 41-42, 130-31. Both editions of Hart and Honorr's book use these
criteria as the bases for the initial distinction between causes and "mere conditions." See id.; H.L.A.
HART & A. HONORt, supra note 10, at 27, 31-33, 38-39, 130-31, 131 n.1. This part of their analysis
corresponds to the tortious-conduct inquiry and is the primary focus of my critique. The second
part of their analysis-the proximate-cause inquiry-is concerned with the problems that arise when
two or more contributing conditions satisfy one of the two criteria. Generally, they treat the last
deliberate intervention or independent abnormal condition as the cause of the injury. Any prior
conditions which also satisfy one of the criteria are reduced to the status of "mere conditions." See
infra note 32.
In the first edition, Hart and Honor6 use the same notion of voluntary intervention in both the
proximate-cause inquiry and the tortious-conduct inquiry. H.L.A. HART & A. HONORt, supra note
10, at 38-39, 129-3 1. In the second edition, however, they use a different definition in the proximate-
cause inquiry. They replace the requirement that the intervenor have intended the actual injury with
a requirement that he have intended to exploit the situation created by the defendant. H.L.A. HART
& T. HONORt, supra note 8, at 136 & n.23. This definition would not work in the tortious-conduct
inquiry, since that inquiry evaluates the conduct of the defendant herself. Moreover, it also appears
to be unworkable in the proximate-cause context, unless it is interpreted to require that the
intervenor be aware of the "untoward" risks involved in the situation created by the defendant and
that the intervenor deliberately (tortiously) act to exploit those risks. See id. at 136. Otherwise, any
voluntary action affecting the situation created by the defendant would be a superseding cause.
In both editions, the notion of abnormal conditions employed in the proximate-cause inquiry is
different than the one employed in the tortious-conduct inquiry. In the proximate-cause inquiry, the
notion of abnormality is expanded to include not only conditions which are per se unusual but also
unusual or coincidental conjunctions of normal events. Conversely, Hart and Honor, state that
intervening human actions do not relieve the defendant of liability unless they are grossly abnormal.
Id. at 77-81, 162-66, 184-85; H.L.A. HART & A. HONORt, supra note 10, at 151-55, 169-70.
The problems created by Hart and Honor&'s attempt to fit the tortious-conduct and proximate-
cause inquiries into a unitary causal rubric are discussed further infra, at text accompanying notes
42-47.
32. H.L.A. HART & T. HONORf, supra note 8, at xlviii-xlix, lxxviii-lxxix, 42-43, 49-50, 71, 74,
77-80, 130-31, 135-36, 162-63.
33. Id. at 41-42, 75-77, 138 & n.40; see supra note 31.
1985] CAUSATION IN TORT LAW 1747

policy judgments on matters of degree and reasonableness. Many of the

circumstances which are treated as rendering even deliberate conduct
nonvoluntary correspond to the recognized moral and legal grounds for
justifying or excusing harmful behavior. Examples include reasonable
actions taken to defend persons or property against reasonably perceived
perils, to safeguard legal rights or interests, to rescue others, and so
forth.3 4 Overall, Hart and Honor6's voluntary-intervention criterion
simply describes one of the principal recognized types of tortious con-
duct: the intentional, unjustified, and unexcused infliction of injury.35
Similarly, the abnormal-condition criterion also calls for nonfactual
and noncausal judgments. When it is applied to human action, it is
essentially identical to the negligence category of tortious conduct, par-
ticularly when attribution of responsibility is at issue. Hart and Honor6
note that the distinction between the abnormal condition and the other
normal conditions is not based on differences in actual contributions to
the consequence, but rather depends, as Malone pointed out, on the con-
text and purpose of the inquiry.3 6 They state that abnormal human con-
duct is conduct that deviates from the usual, expected, or established
standards of behavior. They treat omissions 4s abnormal conditions only
when the failure to act constitutes a breach of duty-that is, when the
omission is negligent.3 7 More generally, they observe that, "[i]n relation
to human conduct, . . . the notion of what is 'natural' is strongly influ-
enced by moral and legal standards of proper conduct, though weight is
also given to the fact that certain conduct is usual or ordinary for a
human being."38
In sum, as Hart and Honor6 acknowledge,39 their two "causal" cri-
teria (voluntary or abnormal conduct) do not have anything to do with
the inquiry into actual contribution to the injury, which is the causal
aspect of responsibility denoted by the word "cause" in the phrase "the
cause." Instead, the two criteria are relevant only in determining
whether the defendant's conduct was "the" (responsible) cause. The two
criteria accomplish this task by focusing on the tortious (intentional or
negligent) character of the conduct that contributed to the injury.
Thus, it is the combination of causation (the causal inquiry), tor-

34. H.L.A. HART & T. HONORt, supra note 8, at Iv,138, 141-60.

35. However, the criterion is too narrow to serve as a complete description of tortious
intentional conduct. See infra text accompanying notes 45-47.
36. H.L.A. HART & T. HONORt, supra note 8, at lxxvi, 11-12, 35-38, 62, 72, 110-11.
37. Id. at 37-38, 50-51, 59-60, 64, 135 n.15, 138-41, 183-85.
38. Id. at 183. The correspondence between the abnormal conduct criterion and negligence
inspired Abraham Harari's effort to demonstrate that negligence (as a supposed aspect of causation)
is the basis of all tort liability. A. HARARI, supra note 13, at 49-55. One of the weaknesses in
Harari's argument is the assumption that the abnormal conduct criterion is a causal criterion.
39. H.L.A. HART & T. HONOR9, supra note 8, at xlviii-xlix, 72-74, 110-11.
 1748 CALIFORNIA LAW REVIEW [Vol. 73:1735

tious conduct by the defendant (the tortious-conduct inquiry), and the

absence of any intervening abnormal physical occurrences or tortious
behavior by third parties (the proximate-cause inquiry) that creates the
strong and usually sufficient case for legal responsibility. Hart and
Honor6's statements treating causation as the primary ground of legal
responsibility are plausible only when it is understood that they are using
the term "causation" to refer to this combination of the three different
elements of tort liability analysis.' They nevertheless insist that their
approach is based almost exclusively on causal principles. This insis-
tence seems to be motivated by a desire to distinguish their common-
sense, principled account of proximate-cause analysis from the social-
policy accounts of the legal realists and others.4 The causal label creates
an appearance of policy-neutrality, which contrasts with the explicit and
broader notion of policy in the alternative accounts.
However, the causal label is misleading and unnecessary, and it cre-
ates major difficulties for Hart and Honor6.
For example, since Hart and Honor6 insist that the defendant's con-
duct was a cause of an injury only if it was deliberate or abnormal (negli-
gent), they are forced to classify strict liability as a noncausal form of
42
liability.
Similarly, because they treat the last deliberate intervention or
independent abnormal occurrence as the cause of an injury, to the exclu-
sion of all prior interventions or occurrences, 43 they find it difficult to
account for cases in which liability is imposed on the defendant despite
some subsequent deliberate intervention or independent abnormal occur-
rence. They describe these as cases of "inducing" or "occasioning" harm
rather than "causing" harm. They claim that there is a "causal connec-
tion" only in a metaphorical sense, even though the defendant's tortious
conduct clearly contributed to the injury. They assert that responsibility
in these cases is based on noncausal policy considerations, and they
therefore use tortious conduct language rather than their usual causal
language to state the grounds for (and limits on) liability. The defendant
is held liable only if he intentionally induced the subsequent deliberate
action of another or negligently provided the opportunity for a subse-
quent deliberate intervention or abnormal occurrence. 44

40. See supra note 12 and accompanying text; see also H.L.A. HART & T. HoNORt, supra note
8, at 117-18, 131-85, 135 n.9, 205-53 (tort liability discussed in terms of causal connection between
tortious conduct and harm, and liability negated despite such causal connection if there was an
intervening abnormal physical occurrence or intervening tortious behavior).
41. H.L.A. HART & T. HONORt, supra note 8, at xxxv-xxxvii, xlvii-lv, 3-7, 24-25, 65-67, 88-
111, 130-32, 254, 304-07.
42. See id. at 85-86.
43. See supra note 32.
44. H.L.A. HART & T. HONORE, supra note 8, at 51-52, 57, 59-61, 71, 81-83, 133, 186. Hart
1985] CAUSATION IN TORT LAW 1749

Finally, Hart and Honor6 face an insuperable dilemma when they

attempt to formulate definitions of their voluntary-intervention and
abnormal-condition criteria that will properly handle both the tortious-
conduct inquiry and the proximate cause inquiry. Since these are sup-
posedly causal criteria, their content should not vary when applied to the
tortious-conduct inquiry as opposed to the proximate-cause inquiry.
Yet, in practice, broader definitions of "voluntary" and "abnormal" are
required for the tortious-conduct inquiry than for the proximate-cause
inquiry. In the proximate-cause inquiry, narrow definitions are needed
to avoid cutting off the defendant's liability when subsequent contribut-
ing conduct, although intentional ("voluntary"), was justified or done in
ignorance of the circumstances or the possible consequences or, although
negligent ("abnormal"), was "routine" rather than "highly unusual." 4 5
In the tortious-conduct inquiry, on the other hand, broader definitions
are needed. The defendant's intentional ("voluntary") intervention often
may be treated as tortious despite ignorance, mistake or necessity, and
his negligent ("abnormal") conduct will be treated as tortious whether it
is "routine" or "highly unusual." 4 6 Hart and Honor6 respond to this
dilemma by adopting a definition of "voluntary" that is suitable for the
proximate-cause inquiry but too narrow for the tortious-conduct inquiry
and, conversely, a definition of "abnormal" that is suitable for the tor-
tious-conduct inquiry but too broad for the proximate-cause inquiry.47
Hart and Honor6 could have avoided these and other difficulties eas-
ily had they correctly described their commonsense principles as princi-
ples of responsibility rather than as principles of causation, and had they
separated the proximate-cause element in these principles from the tor-
tious-conduct and causal elements. As with Malone's argument, Hart

and Honor6's position on this issue is influenced by their view that volitional human actions, unlike
physical events, are neither subject to nor explainable in terms of causal generalizations. Id. at 51-
52, 55-57, 60-61; accord Kadish, Complicity, Cause and Blame: A Study of the Interpretationof
Doctrine, 73 CALIF. L. REV. 323, 326-27, 332-35 & n.8 (1985); Williams, supra note 21, at 66-68.
But, with human actions just as with physical events, if all the relevant circumstances (accumulated
experience and knowledge, mood, and so forth) were the same, surely the decision or action also
would be the same. To assert otherwise is to assert that human action is random or arbitrary. Cf
H.L.A. HART & T. HONOR-, supra note 8, at 56-57 (acknowledging that generalizations apply to
human action, but nevertheless denying repeatability in identical circumstances). Human action is
less regular and predictable than physical events because humans learn from prior experiences and
new information, because the range of relevant conditions is much broader, and because the
applicable causal generalizations are much more complex and less well understood. See generallyT.
BEAUCHAMP & A. ROSENBERG, supra note 21, at 314-27; J. MACKIE, THE CEMENT OF THE
UNIVERSE 120-26 (1974); Moore, Causation and the Excuses, 73 CALIF. L. REV. 1091, 1112, 1124-
27, 1132-37 (1985).
45. H.L.A. HART & T. HONORt, supra note 8, at 41-42, 75-77, 136, 152-53, 183-84; W.
PROSSER, supra note 1, § 44, at 274, 276-78, 282.
46. W. PROSSER, supra note 1, §§ 7-8, at 30-33; id. § 9, at 35; id, § 17, at 99-100; id. § 24, at
126-27; id. § 31, at 145; id. § 32, at 150-51.
47. See supra note 31 and text accompanying notes 33-38.
1750 CALIFORNIA LAW REVIEW [Vol. 73:1735

and Honor6's theory illustrates the conceptual confusion and practical

difficulties that inevitably result when these three elements of tort liabil-
ity analysis are not carefully distinguished from one another.
3. Epstein
Hart and Honor6's book has inspired several other attempts to
develop causal theories of tort liability."a The best known example in the
United States is Richard Epstein's theory of "strict" liability.4 9
Epstein, even more than Hart and Honor6, opposes the ad hoc,
social-policy accounts of tort liability, including the currently fashionable
wealth-maximization accounts. 50 He argues that tort liability should be
based on principles of individual autonomy and responsibility, with the
dividing line between autonomy and responsibility being marked by cau-
sation of harm to others. 1 Epstein's emphasis on individual autonomy is
coupled with a strong preference for precise rules of responsibility. He
seeks to eliminate or greatly restrict those aspects of traditional tort lia-
bility analysis-such as tests of reasonableness, negligence, or proximate
cause-which, because of their imprecision, can be used by judges or
juries to validate substantial incursions on individual autonomy. 2
Although Epstein refers to his theory as a theory of strict liability, it
is in fact a system of defeasible absolute liability. Tort theories are classi-
fied as "absolute," "strict," or "fault-based" depending on the elements
required to state a minimally sufficient prima facie case.5 3 Traditionally,
absolute liability has meant prima facie liability based merely on causa-
48. See supra note 13. Abraham Harari's causal theory is discussed briefly supra in note 38.
John Borgo's analogical reasoning approach suffers from precisely the same flaw that Borgo
correctly identifies in Epstein's approach. It abstracts responsibility paradigms from situations while
overlooking critical aspects of those situations and then tries to apply the abstracted paradigms to
situations where the critical aspects are different. See Borgo, supra note 11, at 429-31, 436-40.
49. Epstein, supra note 13; Epstein, Defenses and Subsequent Pleas in a System of Strict
Liability, 3 J. LEGAL STUD. 165 (1974) [hereinafter cited as Epstein, Defenses]; Epstein, Intentional
Harms, 4 J. LEGAL STUD. 391 (1975) [hereinafter cited as Epstein, Intentional Harms]; Epstein,
Nuisance Law: Corrective Justice and its Utilitarian Constraints, 8 J. LEGAL STUD. 49 (1979)
[hereinafter cited as Epstein, Nuisance Law]; Epstein, Causation and Corrective Justice: A Reply to
Two Critics, 8 J. LEGAL STUD. 477 (1979) [hereinafter cited as Epstein, Reply]. Hart and Honor6's
influence on Epstein is indicated in, for example, Epstein, supra note 13, at 161-64 & n.39; Epstein,
Intentional Harms, supra, at 431-32 & n.100; Epstein, Reply, supra, at 479, 486 n.33. Compare
Epstein, supra note 13, at 166-69, 175 (building on simple transitive propositions), with H.L.A.
HART & A. HONORf, supra note 10, at 26-28, 68-69 (same). However, in their second edition Hart
and HonorE reject Epstein's "causal maximalism." H.L.A. HART & T. HONORI, supra note 8, at
lxxiii-lxxvii; see id. at 30-31, 96-97 (notion of force is inadequate basis for causal theory).
50. Epstein, supra note 13, at 151-57, 198-99.
51. Id. at 198-201, 203-04; Epstein, Intentional Harms, supra note 49, at 441-42; Epstein,
Reply, supra note 49, at 479-80.
52. Epstein, supra note 13, at 198-99; see id. at 162-66, 185-87; Epstein, Defenses, supra note
49, at 178-81, 184; Epstein, Reply, supra note 49, at 478-80 & n.5.
53. See Epstein, Pleadings and Presumptions,40 U. CHI. L. REv. 556, 558-59, 570-71, 577
(1973) (minimally sufficient elements for responsibility constitute prima facie case and establish basic
1985] CAUSATION IN TORT LAW

tion of harm to another. 4 Strict liability requires, in addition to causa-

tion, that the harm result from conduct which created a foreseeable,
unaccepted risk of harm to another that could have been avoided.5 5
Fault liability requires, in addition to causation, that the harm result
from conduct which should have been avoided given the foreseeable
risks.5 6 Epstein purports to base prima facie liability simply on causation
of harm to another.57 This prima facie absolute liability, however, may
be eliminated or reduced if certain defenses are established. 58
Epstein's normative arguments on behalf of his absolute liability
theory assume that there are only two possible types of prima facie tort
liability-fault and absolute. 9 His negative arguments point out the
flaws in the fault theories and the failure of those theories to account for
various cases in which liability was imposed (or arguably should have
been imposed) although there was no faulty behavior by the defendant. 6"
The cases he discusses, however, are strict liability cases-cases in which
the defendant knew or should have known that he was imposing an unac-
cepted risk on others.6 1 Similarly, his affirmative arguments are almost

theory of recovery, even when responsibility defeasible through defenses or other subsequent
pleadings).
54. E.g., O.W. HOLMES, supra note 1, at 67-68, 72; Ehrenzweig, Negligence Without Fault, 54
CALIF. L. REv. 1422, 1426 nn.13 & 18 (1966), and the sources cited therein.
55. Eg., PROSSER & KEETON, supra note 7, § 75, at 536-38; id. § 78, at 555-56; Ehrenzweig,
supra note 54, at 1450-51 & n.l; see O.W. HOLMES, supra note 1, at 77, 93-95, 115-18.
56. E.g., O.W.HOLMES, supra note 1, at 85-88; Ehrenzweig, supra note 54, at 1444-45.
57. Epstein, supra note 13, at 152, 168-69.
58. These defenses include reciprocal causation of harm to the defendant by the plaintiff,
assumption of the risk by the plaintiff, and trespass on the defendant's property by the plaintiff.
Epstein, Defenses, supra note 49, at 167-68, 174, 185, 201.
59. Epstein, supranote 13, at 152, 171; Epstein, IntentionalHarms, supranote 49, at 398. This
erroneous assumption also appears in most economic analyses of tort liability. See, eg., Brown,
Toward An Economic Theory of Liability, 2 J. LEGAL STUD. 323 (1973); Shavell, Strict Liability
Versus Negligence, 9 J.LEGAL STUD. 1 (1980).
60. Epstein, supra note 13, at 153-60, 169-71.
61. Morris v. Platt, 32 Conn. 75 (1864); Vincent v. Lake Erie Transp. Co., 109 Minn. 456, 124
N.W. 221 (1910). Both cases are discussed in Epstein, supra note 13, at 157-60. See also Bolton v.
Stone, 1951 A.C. 850, rev'g [1950] 1 K.B. 201 (C.A. 1949), rev'g [1949] 1 All. E.R. 237 (Manchester
Assizes 1948). Bolton is discussed in Epstein, supra note 13, at 169-71. In Bolton, the defendant
cricket club maintained and used a playing field not only with the knowledge that cricket balls might
be hit out of the grounds and endanger passersby such as the plaintiff, but also with the knowledge
that one of the objects of the game is to hit the ball away from the opposing team, the further away
the better. Although balls are rarely hit out due to the size of the playing fields, those that are hit out
earn extra points. See 1 THE ENCYCLOPEDIA OF SPORT 447 (1898). Thus, although the risks to
passersby such as the plaintiff were slight, the possibility was knowingly and eagerly pursued as an
integral part of the defendant's activity. This latter aspect of the case, together with the likely
severity of the harm should a ball actually escape the grounds and hit someone, may well explain (a)
the House of Lords' unease in refusing to hold the defendant liable, (b) the public interest in the case,
and (c) the decision of the Cricket Clubs of England to compensate the plaintiff despite the lack of a
holding of legal liability. See Epstein, supra note 13, at 170. It seems unlikely that the case would
have aroused such controversy if the possibility that cricket balls would escape the grounds had been
completely unintended, unknown, and unforeseeable.
1752 CALIFORNIA LAW REVIEW [Vol. 73:1735

always couched in strict-liability terms. They explicitly or implicitly rely

not only on causation of harm by the defendant, but also on the defend-
ant's acting despite actual or constructive knowledge of a foreseeable,
unaccepted risk to another. 2
This mental element, however, is absent in Epstein's description of
the prima facie case under his theory, which initially is based solely on
causation of harm. Epstein does not provide any general definition or
account of causation. He rejects the traditional "but for" (necessary con-
dition) test because it treats too many conditions as causes of the injury
and therefore allegedly throws all the critical liability decisions into the
policy-laden proximate-cause inquiry.63 In place of the but-for test and
the proximate-cause inquiry, he identifies a set of "causal paradigms"
which purportedly explicate the concept of causation and constitute nec-
essary and prima facie sufficient bases for the imposition of liability."
But neither of these claims holds up. Epstein's paradigms do not
explicate the concept of causation.6 5 Instead, he uses the concept of cau-
sation, undefined and unelaborated, to construct what are actually para-
digms of responsibility, by combining certain methods of causing harm
(primarily force and the threat of force) with certain noncausal criteria.
However, the prescribed methods of causing harm are too restrictive,
while the prescribed noncausal criteria do not include the mental element
required for true strict liability. Consequently, Epstein's theory produces
results which diverge substantially from those reached by the courts.66
The first paradigm is "Ahit B," which is meant to be an abbreviated
way of saying that A, by application of force to B or B's property, broke,
cut, or otherwise harmed B or B's property.67 Note the implicit circular
reference to the concept of causation in this "causal" paradigm. The

62. Epstein, supra note 13, at 158-59 & n.22, 173; Epstein, Defenses, supra note 49, at 169, 213;
Epstein, Intentional Harms, supra note 49, at 398, 414.
63. Epstein, supra note 13, at 161-63; Epstein, Reply, supra note 49, at 478-79. But see Epstein,
supra note 13, at 183 (but-for test applied to dangerous position of vase), 184 (same: dangerous
condition of car).
64. Epstein, supra note 13, at 160, 165-66, 168-69, 190-91, 194-95; Epstein, Defenses, supra
note 49, at 167-68; Epstein, Intentional Harms, supra note 49, at 398-400; Epstein, Reply, supra note
49, at 479-80.
65. This point has also been made by Borgo, supra note 11, at 427-31; see supra note 48.
66. Epstein states that his theory is predominantly normative. Epstein, supra note 13, at 151;
Epstein, Reply, supra note 49, at 477. Nevertheless, he recognizes that any plausible normative
theory cannot stray too often or too far in its results from the commonly accepted notions of
responsibility embodied in traditional tort law, and he claims to base his theory on those commonly
accepted notions. See Epstein, supra note 13, at 151, 163-64, 166; Epstein, Reply, supra, at 479, 503-
04. Indeed, he argues that his theory best accounts for the results reached by the courts. Epstein,
supra note 13, at 165-66, 189; Epstein, Defenses, supra note 49, at 168-69, 173-74, 204-05, 213-15;
Epstein, Intentional Harms, supra note 49, at 406-07; Epstein, Reply, supra, at 480.
67. Epstein, supra note 13, at 166-67; Epstein, Intentional Harms, supra note 49, at 399;
Epstein, Reply, supra note 49, at 480-81.
1985] CAUSATION IN TORT LAW 1753

paradigm implicitly states that A, by application of force, caused harm to

B or B's property-for example, that the force initiated (caused) by A
caused B's cut, which is itself the harm for which compensation is
sought or which led to (caused) the ultimate harm for which compensa-
tion is sought. The paradigm does not explain or help to identify any of
these instances of causation. It rather assumes that we somehow per-
ceive them, and it restricts liability to those that involve force as the
causal mechanism.
Epstein also requires that the force have been initiated by a voli-
tional act of A, rather than by someone's throwing A or by some auto-
matic or reflex action ofA. 68 Thus the paradigm includes, in addition to
(unelaborated) actual causation, the noncausal criteria of a volitional act
and a particular method of causation (the application of force).
Similarly, the second paradigm, "A frightened B," requires, in addi-
tion to actual causation, that A have caused the harm to B by a volitional
act and through the causal mechanism of fright or shock. As originally
stated, the paradigm would cover cases in which extrasensitive or even
normal people are frightened by the completely innocent and inoffensive
acts of others. Epstein suggests that, in these cases, the frightened person
should be treated as having caused her own fright, but he does not indi-
cate which of his causal paradigms would apply.6 9 None would, espe-
cially since the person's fright is not considered to be volitional.
Eventually Epstein modifies the paradigm by requiring that A's volitional
act be an "offer of force" against B, or, to the same effect, that a de
minimus rule be applied to A's conduct.7" These modifications, of
course, introduce additional noncausal criteria into the paradigm.
The third paradigm holds A liable to C if "A [by a volitional act]
compelled B to hit C."7 1 The compulsion must be accomplished through
force or the threat of force.7" Under this third paradigm, B as well as A
is liable to C, since B satisfies the initial paradigm, "B hit C." Moreover,
B has an action against A for the damages that he is required to pay C,
since A satisfies the paradigm, "A hit (or offered to hit) B."7 3
The fourth paradigm most clearly demonstrates the circularity and
ambiguity of Epstein's causal paradigms. It holds A liable to B if "A [by
a volitional act] created [caused] the dangerous condition that resulted in

68. Epstein, supra note 13, at 166-67. See generally Moore, supra note 44, at 1132-36
(distinguishing action from mere causal sequence).
69. Epstein, supra note 13, at 172.
70. Epstein, Reply, supra note 49, at 483-84.
71. Epstein, supra note 13, at 174. See generally Moore, supra note 44, at 1129-32
(distinguishing compulsion from mere causation).
72. See Epstein, Defenses, supra note 49, at 175; Epstein, IntentionalHarms, supranote 49, at
399.
73. Epstein, supra note 13, at 175; Epstein, Defenses, supra note 49, at 174-75.
1754 CALIFORNIA LAW REVIEW [Vol. 73:1735

[caused] harm to B." 74 Note that two instances of causation are implied
by the expression "resulted in." First, some act or event must "trigger"
the dangerous condition, and, second, the triggered condition must then
cause the harm.
Epstein attempts to avoid the charge that this paradigm, by use of
the expression "resulted in" (he overlooks the verb "created"), defines
causation in terms of itself. He says that the expression "resulted in"
must be replaced in each particular case with a description of the act or
event which is the immediate cause of the harm, using one of the three
previously elaborated paradigms-force, fright or compulsion.75 How-
ever, even if both the expression "resulted in" and the verb "created"
were interpreted and used in this restricted fashion, it would only push
the circularity one level deeper, since-as we have seen-the three previ-
ous paradigms themselves contain implicit, unelaborated causal state-
ments. Moreover, Epstein does not actually restrict either term in this
fashion.
Thus, in Epstein's examples dangerous conditions are "created" by
"storing" or "handling" explosives, by "placing" or "leaving" an object
in a precarious or obstructive position, or by "manufacturing" a defective
product. And dangerous conditions are "triggered" by a "small change
in conditions, like temperature and humidity," or by mere use of a defec-
tive product (which might simply be turning it on or eating it).7 6 Clearly,
given these examples, the creation or triggering of dangerous conditions
is not limited to the use or threat of force, unless "force" is construed so
broadly as to encompass any action.
Similarly, in Epstein's examples dangerous conditions, when trig-
gered, can cause harm not just by "releas[ing] or otherwise set[ting] in
motion large forces" that are associated with the dangerous condition
itself, but also by absorbing or "redirecting" forces that are not associ-
ated with the dangerous condition. For example, an obstruction in a
highway is treated as a dangerous condition that "results in" injury to B
when B drives into it or swerves to avoid it." Here "results in" cannot
be replaced by one of the three previous paradigms: the obstruction did
not hit B, or offer to hit B, or compel anyone to hit B. Rather, the con-
verse is true: the obstruction was hit by B or, in the swerve case,
"offered" to be hit by B!78

74. Epstein, supra note 13, at 177.

75. Id.
76. Id. at 177-78.
77. Id. at 178, 185.
78. Epstein might argue that A's obstruction compelled B to hit A's obstruction. This
argument, however, would have to rely on a definition of compulsion that includes more than the
actual or threatened use of force by A. At most, A only threatened to absorb B's force. Moreover,
the argument would only establish B's right to be indemnified by A for damages that B caused to A's
1985] CAUSATION IN TORT LAW 1755

In other examples, the triggered condition "results in" harm

through causal mechanisms involving fire, pollution, unwanted oxygen,
delays, and so forth, in which there is no use or offer of force by anyone,
unless once again "force" is interpreted so broadly that it applies to
anything.79
In addition to its causal circularity and ambiguity, the dangerous-
condition paradigm contains the most obvious noncausal criterion of any
of the paradigms: the requirement that the condition which resulted in
the harm have been "dangerous." Epstein tries to tie the term "danger-
ous" into his notion of causation-as-force by limiting the paradigm to
those "dangerous conditions that release or otherwise redirect forces in
the narrowest sense of causation."" 0 Thus, he asserts, a sharp knife is a
"mere" condition rather than a "dangerous" condition, because it con-
tains no store of energy waiting to be released (nor, presumably, does it
"redirect" forces)."1 But, as we have already seen, Epstein does not
adhere to this limitation. An obstruction, when hit by something, does
not release or redirect forces. It absorbs forces, as does B in the para-
digm "A hit B." Similarly, if a knife with a sharp edge does not release
or redirect forces, then neither does a toy that is defective because it has a
sharp edge. Yet these two examples clearly illustrate two of Epstein's
three categories of dangerous conditions: things in dangerous positions
and defective products. (The third category is inherently dangerous or
ultrahazardous things, such as stored explosives.) 2
The examples discussed so far indicate the accordion-like quality of
Epstein's paradigms and his causal theory as a whole, which alternately
expand and contract to fit the particular argument. When Epstein wants
to deny liability in a certain case or to emphasize the coherence of his
theory, he insists on causation through force, the offer of force, or the
creation of a dangerous condition which releases or redirects force.8 3 But
strict adherence to this limitation would require him to deny liability in
many types of cases in which liability is imposed by the courts: fires,
poisonings, fraud, defamation, emotional distress cases not involving the
use of force, and so forth. When discussing these types of cases, Epstein

obstruction (or other object, in the swerve case). There is no embedded paradigm, "A hit (or offered
to hit) B" that would support an action by B against A for damages suffered by B when B hit the
obstruction. See supra text accompanying notes 71-73. Nor, unless we are to twist language, can it
be said that A compelled B to hit himself. B hit the obstruction (or other object), not himself.
79. Epstein, Nuisance Law, supra note 49, at 56-57, 68, 100; Epstein, Reply, supra note 49, at
486-87.
80. Epstein, supra note 13, at 185; see id. at 179.
81. Id. at 179.
82. See id. at 177-78; see also Epstein, IntentionalHarms, supra note 49, at 415 (barbed wire as
a dangerous condition).
83. Epstein, supra note 13, at 180, 185; Epstein, IntentionalHarms, supra note 49, at 399-400,
431; Epstein, Reply, supra note 49, at 479-81.
1756 CALIFORNIA LAW REVIEW [Vol. 73:1735

usually abandons the force limitation and recognizes any "paradigm"

that expresses a "nonreciprocal" (transitive) causal relationship-for
example, "A poisoned B," "A segregated B where aid could not be
given," "A misled B," "A entered B's land," "A damaged B's property,"
84
or "A infringed B's interest."
At this point, the alleged power of the causal paradigms disappears.
As Epstein implicitly admits, the inquiry has dissolved into a debate over
rights and protected interests.8 He belatedly notes that the question of
nonreciprocal causation of harm-of who harmed whom-cannot be
86
resolved without a prior determination of the parties' respective rights.
Indeed, he goes so far as to identify the "causal question" with "the ques-
tion of what legal right of the plaintiff was involved." 87 Unfortunately,
he assumes that a person's rights in her body or property are absolute:
theoretically, any invasion creates liability.88 As a result, he is forced to
introduce "utilitarian constraints" into his system to avoid liability for de
minimus harms.89
Similar problems appear in Epstein's infrequent discussions of prox-
imate-cause limitations. Although Epstein intended to replace tradi-
tional proximate-cause analysis with his causal paradigms,90 he never
demonstrates how his paradigms accomplish this goal, and it is not clear
how they could accomplish it. The four initial paradigms merely require
that the injury result from force (or an offer of force) initiated by A him-
self or by someone compelled by A or by a dangerous condition created
by A. The basic proximate cause problems remain. How far is the force

84. Epstein, supra note 13, at 195 n.104; Epstein, Defenses, supra note 49, at 202; Epstein,
Intentional Harms, supra note 49, at 432; Epstein, Nuisance Law, supra note 49, at 51. In the
emotional distress cases not involving the use or threat of force, Epstein adheres to the force
limitation and denies liability under his "invasion model." He suggests that liability in these cases is
based on "breaches of implied contracts for good social behavior" and that the "contractual mold" is
appropriate since these cases "all involve direct personal contact between plaintiff and defendant."
Epstein, Nuisance Law, supra, at 64 n.44. This reasoning, of course, would apply to almost every
tort. Epstein's discussion demonstrates the difficulties created by his narrow view of tort theory and
his frequent resort to dubious contractual arguments to escape those difficulties. See also Epstein,
Reply, supra note 49, at 481 (drivers on highway are not strangers since they are "bound together by
. . . statutory. . . bonds," thus calling for "complicated notions of 'avoidance of harm' which are
not captured by my paradigms").
85. Epstein, IntentionalHarms, supra note 49, at 432-41.
86. Id. at 432-34, 441; Epstein, Nuisance Law, supra note 49, at 50-53. Compare Epstein, supra
note 13, at 164-69 (nonreciprocal causation defines rights), with Epstein, Nuisance Law, supra, at 58-
60 (question of causation depends on prior definition of rights) and Epstein, Reply, supranote 49, at
498 & n.64 (references to Epstein's gradual recognition of the primacy of rights analysis).
87. Epstein, IntentionalHarms, supra note 49, at 433.
88. See Epstein, Nuisance Law, supra note 49, at 63-64 & n.40, 82, 92-93; Epstein, Reply, supra
note 49, at 488-500.
89. Epstein, Nuisance Law, supra note 49, at 50, 74-102; Epstein, Reply, supra note 49, at 483-
84, 488-89, 503.
90. Epstein, supra note 13, at 161-63, 164-65, 168, 184-85; Epstein, Reply. supra note 49, at
478-80.
1985] CAUSATION IN TORT LAW 1757

to be traced? Is A liable for harm caused by the dangerous condition

despite intervening events?
Epstein compares the case in which B is injured when he swerves to
avoid A's obstruction with the case in which B is injured while traveling
back and forth to remove A's obstruction. Epstein claims that the dan-
gerous-condition paradigm distinguishes the two cases. A is liable only in
the swerve case, because in that case A's obstruction "redirected" B's
force, and B was injured by the redirected force. But this is true also in
the traveling-back-and-forth case. Epstein's further assertion that the
"causal connection ends once the road is repaired" is unsupported by his
paradigms. Moreover, the logic of the assertion would permit (indeed,
require) liability if the injury occurred on the way to the obstruction.9"
Epstein sometimes states that the injury must be produced directly
and immediately by the force,92 but he provides very little elaboration.
In some cases, he is willing to treat intervening events as "means," 9 3
while in other cases he is willing to trace forces through lengthy, complex
causal chains. 94 On the few occasions that he discusses proximate-cause
problems, he frequently relies on traditional proximate-cause
arguments. 95
In sum, Epstein's causal paradigms as initially elaborated are much
too narrow to serve as the basis for a comprehensive theory of tort liabil-
ity. When he expands the content and number of the paradigms to make
his theory more plausible, the paradigms become superfluous. There is
liability for any volitional act that infringes a legally protected interest,
and any method of causation qualifies as an infringement, subject to
unspecified but apparently traditional proximate-cause limitations.
However, no matter how many methods of causation are allowed,
Epstein's theory still diverges substantially from the decisions of the
courts, since his paradigms do not include the critical mental element
that is required for liability. A volitional act is not enough. At a mini-
mum, the courts require the plaintiff to establish that the defendant knew
or should have known that his act would impose an unaccepted risk on
others. 96 Epstein's paradigms do not include this element (except per-
haps the fright paradigm when modified to require an "offer" of force).
91. Epstein, supra note 13, at 185.
92. Id. at 177, 184-185; Epstein, Intentional Harms, supra note 49, at 399-400; Epstein,
Nuisance Law, supra note 49, at 56.
93. Epstein, supra note 13, at 167-68.
94. Epstein, Nuisance Law, supra note 49, at 56-57 (nuisance cases); see also Epstein,
Intentional Harms, supra note 49, at 429 (defendant used force to drive away plaintiff's potential
customers).
95. See Epstein, supra note 13, at 174 n.70, 176 n.73, 185; Epstein, Defenses, supra note 49, at
183-84; Epstein, IntentionalHarms, supra note 49, at 431-32 & n. 100; Epstein, Nuisance Law, supra
note 49, at 56-57; Epstein, Reply, supra note 49, at 486-87 & n.37.
96. O.W. HOLMES, supra note 1, at 76-78; see also id. at 92-95, 115-18.
1758 CALIFORNIA LAW REVIEW [V/ol. 73:1735

The gap between Epstein's theory and the actual practice of the
courts is narrowed somewhat in the reciprocal-causation cases. In these
cases, both the plaintiff and the defendant have harmed each other under
one of the causal paradigms. The plaintiff's reciprocal causation of harm
to the defendant provides a defense, in the second stage of the pleadings,
to the plaintiff's prima facie case against the defendant, which was estab-
lished in the first stage. Epstein then allows the plaintiff, in a third stage
of the pleadings, to reestablish his claim by proving that the defendant's
conduct was tortious.97 But by then it is too little and too late. The
tortious-conduct inquiry is never reached in the many cases in which the
pleadings do not progress beyond the first stage because there is no recip-
rocal causation. Even when the tortious-conduct inquiry is reached,
Epstein refuses to consider one of the principal types of tortious con-
duct-negligence-because of his strong distrust of imprecise liability
principles.9" Thus, under Epstein's theory, a driver whose car acciden-
tally breaks down on the highway and who sets out adequate warnings
for other drivers is liable to, but cannot recover from, the operator of a
truck who negligently crashes into the car. The driver of the car creates
a dangerous-condition by blocking the right of way, the truck's crashing
into the car merely completes the paradigm, and the truck operator's
negligence is (according to Epstein) immaterial. 99
It should by now be clear that causation is not equivalent to respon-
sibility, nor is it prima facie sufficient for liability. The tortious-conduct
and proximate-cause inquiries are distinct from the causal inquiry, and
all three are required to complete the liability analysis under traditional
notions of legal responsibility.
An important question remains, however. Is the causal inquiry a
significant part of tort analysis, or is it, as alleged by Malone and feared
by Hart, Honor6, and Epstein, dominated by the tortious-conduct and
proximate-cause inquiries?

97. Epstein, Defenses, supra note 49, at 177-84; Epstein, Intentional Harms, supra note 49,
passim.
98. Epstein, Defenses, supra note 49, at 172; Epstein, Reply, supra note 49, at 482; see supra
note 52 and accompanying text. But cf Epstein, Defenses, supra,at 177-84 & n.49 (statutory duties);
Epstein, Nuisance Law, supra note 49, at 71-72 & n.58 (reasonableness considerations in highway
cases); Schwartz, The Vitality of Negligence and the Ethics of Strict Liability, 15 GA. L. REV. 963,
986-89, 994 (1981) (discussing Epstein's reliance on negligence notions in certain situations); supra
note 84 (reasonableness considerations in emotional distress cases).
99. Epstein, supra note 13, at 180-81, 191-92 & n.99; Epstein, Defenses, supra note 49, at 177-
1985] CAUSATION IN TORT LAW 1759

B. The Significance of the CausalInquiry: Linking the Causal

Inquiry and the Tortious-ConductInquiry Through the
Tortious-Aspect CausationRequirement
The significance of the causal inquiry in tort law varies dramatically
depending on how it is linked to the tortious-conduct inquiry. There are
two principal options. The first is to apply the causal inquiry to the
defendant's conduct as a whole, for example, operating a hotel or driving
a car. The second is to focus the causal inquiry on the tortious aspect of
the defendant's conduct, for example, failure to provide a fire escape in
the hotel, or excess speed or inattentiveness while driving the car. The
choice between these two approaches is a decision of policy or principle.
This, however, does not make the causal inquiry itself any less factual.
Adoption of the first (overall-conduct) approach makes it easy to
satisfy the actual-causation requirement, unless the defendant had no
connection with any of the conditions or events that contributed to the
injury. The causal inquiry therefore plays a relatively minor role in the
determination of liability. Instead, liability turns primarily on policies
related to deterrence of tortious behavior, allocation of losses to a tor-
tious defendant when the losses are "within the risk" created by the tor-
tious behavior (whether or not they were caused by the tortious aspect of
the behavior), providing incentives for wealth-maximizing behavior, and
so forth.100
Adoption of the second (tortious-aspect) approach, however, signifi-
cantly constrains pursuit of such policy goals by according a much
greater role to the causal inquiry in determining tort liability. The
defendant cannot be held liable unless the tortious aspect of his conduct
contributed to the injury. 1M The role of the causal inquiry becomes even
more significant if, as appears to be the case, defendants are often held
liable for harm caused by the tortious aspect of their behavior even
though the harm was neither foreseeable nor "within the risk."' 0 2
It is clear from the cases that the courts follow the tortious-aspect
approach.103 In fact, it is so clear that many readers may wonder why I
have devoted even this many lines to belaboring the obvious. My justifi-
cation is that, surprisingly, very few legal writers have noted this point,

100. E.g., G. CALABRESI, THE COSTS OF ACCIDENTS 6-7 & n.8, 131-97 (student ed. 1970);
Calabresi, supranote 9; Green, supranote 8; Pound, supra note 7; Thode, The Indefensible Use of the
Hypothetical Case to Determine Cause in Fact, 46 TEx. L. REV. 423 (1968).
101. For a discussion of the difficulties that this restriction creates for the wealth-maximizing
theories of tort liability, see Wright, supra note 9.
102. See infra text accompanying notes 161-64.
103. E.g., Weeks v. McNulty, 101 Tenn. 495, 48 S.W. 809 (1898); Texas & Pac. Ry. v.
McCleery, 418 S.W.2d 494 (Tex. 1967); Peterson v. Nielsen, 9 Utah 2d 302, 343 P.2d 731 (1959);
Sowles v. Moore, 65 Vt. 322, 26 A. 629 (1893); see RESTATEMENT (SECOND) OF TORTS § 432
(1965); RESTATEMENT OF TORTS § 519 (1938).
1760 CALIFORNIA LAW REVIEW [Vol. 73:1735

much less emphasized its significance. 'I A substantial number of writ-

ers have expressly rejected the tortious-aspect approach to the cause-in-
fact inquiry."0 5 Others have written ambiguously about the need to
establish a causal connection between the defendant's tortious conduct
and the injury. 106
The proponents of the overall-conduct approach replace the tor-
tious-aspect causation requirement with risk-related limitations on liabil-
ity. One theory, championed by Leon Green, relies on ad hoe, open-
ended policy considerations as part of the duty (tortious-conduct)
inquiry. The other theory, developed most extensively by Robert Kee-
ton, limits liability to harm that is "within the risks" that made the
defendant's conduct tortious. This limitation occurs as part of either the
tortious-conduct inquiry or the proximate-cause inquiry.
In this section I first argue that the risk theorists must smuggle the
tortious-aspect causation requirement into their risk analyses to obtain

104. Those who have at least noted the point include A. BECHT & F. MILLER, supra note 8,
passim; Carpenter, Workable Rules for Determining Proximate Cause (pt. 2), 20 CALIF. L. REV.
396, 408-19 (1932); Phillips, Reflections on Factual Causation, 1978 WASH. U.L.Q. 661, 663;
Williams, supra note 21, at 64-65, 70-71, 73-75; see also H.L.A. HART & T. HONORt, supra note 8,
at xxxvii-xxxviii, lviii-lxi, 117-21, 208-11, 293-94; R. KEETON, LEGAL CAUSE IN THE LAW OF
TORTS 4-9, 12-13, 18 (1963). The economic analysts generally have assumed the tortious-aspect
approach when they address the causation issue, but many have assumed the overall-conduct
approach in their articles on tortious conduct. See Brown, supra note 59, at 327-29, 333-34 (overall
conduct); Calabresi, supra note 9, at 79, 85-86 (tortious aspect); Calfee & Craswell, Some Effects of
Uncertainty on Compliance with Legal Standards, 70 VA. L. REV. 965, 975, 990 & n.56 (1984)
(overall conduct); Grady, supra note 9, at 804, 814-15, 824-25 (tortious aspect); Landes & Posner,
supra note 9, at 111-18 (tortious aspect); Landes & Posner, The Positive Economic Theory of Tort
Law, 15 GA. L. REv. 851, 868-70, 880-82 (1981) (overall conduct); Shavell, supra note 9, at 463-64,
481-82, 489 (tortious aspect); Shavell, supra note 59, at 10, 11 n.18, 15 (overall conduct).
105. 2 F. HARPER & F. JAMES, supra note 2, § 20.5, at 1138 & n.17 (But cf id. § 20.2 n.4
comment, at 93-94 (Supp. 1968)); R. KEETON, supranote 104, at viii-ix, 10-16; Cole, supra note 7, at
459-60, 473-75, 482-86 & n.80, 491, 498-507 & n.133; Crowe, The Anatomy ofa Tort-Greenian, as
Interpreted by Crowe, Who Has Been Influenced by Malone-A Primer, 22 Loy. L. REV. 903, 904-
05, 920-22 (1976); Epstein, supra note 13, at 152, 168-69, 181; Green, supra note 8, at 544 & n.3,
546-52, 562, 569 n.77, 576; Henderson, A Defense of the Use of the Hypothetical Case to Resolve the
Causation Issue-The Need for an Expanded, Rather than a Contracted, Analysis, 47 TEx. L. REV.
183, 185, 195-98, 200-01, 212-13 (1969); Morris, supra note 3, at 1096-97, 1104; Thode, supra note
100, at 424-25, 428-29; see also Pedrick, supra note 7, at 656-57 & n.41.
Although Robert Keeton claims to be using the tortious-aspect approach, he substitutes the
"harm within the risk" approach to proximate-cause limits for the tortious-aspect approach to
cause-in-fact. See infra text accompanying notes 118-26. Keeton and Thode each report Warren
Seavey's rejection of the tortious-aspect approach. R. KEETON, supra note 104, at 12-13; Thode,
supra note 100, at 429 & n.23. Pedrick cites Prosser as also supporting the overall-conduct
approach, but the cited portion of Prosser's text uses the defendant's negligence (excess speed) rather
than his conduct as a whole (driving) as the basis for the causal inquiry. Pedrick, supra note 7, at
656 n.41 (citing W. PROSSER, supra note 1, § 41, at 237 & n.6).
106. E.g., J. FLEMING, supra note 4, at 170-71; C. MORRIS & C.R. MORRIS, supra note 4, at
154-201; W. PROSSER, supra note 1, § 30, at 143; id. § 41, at 236-43; id. § 42, at 249; id. § 45, at 289;
PROSSER & KEETON, supra note 7, §§ 41-45, at 263-321; Smith, Legal Cause in Actions of Tort (pts.
1-3), 25 HARV. L. REV. 103, 223, 303 (1911-1912).
1985] CAUSATION IN TORT LAW 1761

results that are consistent with the decisions of the courts. I then elabo-
rate the tortious-aspect causation requirement and apply it to a broad
range of cases to demonstrate its importance in determining liability.

L Green's Duty-Risk Theory

Leon Green and his followers insist that the causal inquiry be
applied to the defendant's conduct as a whole. They vigorously reject
any requirement of a causal relation between the injury and the tortious
aspect of the defendant's conduct. For them, the critical factors in deter-
mining liability are the relevant social policies. They claim that the court
should weigh these policies in each case to decide whether the injury
represents the type of risk for which the defendant should be held
responsible (the duty issue). Furthermore, these policies should also be
weighed by the jury, in conjunction with the other contributing factors
and the foreseeability of harm of the sort that occurred, to decide
whether the defendant should compensate the plaintiff (the negligence, or
breach of duty, issue). 10 7
Green offers the following example to demonstrate how his
approach "relieves the causation concept of most of its burdens and con-
fusing terminology."'' 0 8 A driver X runs into P without trying to use his
brakes. X was going too fast to stop anyway. Moreover, unbeknown to
X, the brakes were defective. D, who knew the brakes were defective,
had sold the car to X without disclosing the defective condition of the
brakes. Green notes that, under his approach, there is a clear causal
connection between each of X's and D's activities and P's injury. "D's
selling the car to X [regardless of D's knowledge of the defective brakes]
contributed to P's injuries.. . . X's buying and driving the car [regard-
less of his speeding and failure to try to use the brakes] contributed to P's
injury. ' ' °0 The liability determination, therefore, turns on the scope of
the duties and the question of whether either D or X was negligent with
respect to P.110
Green states without elaboration that "X owed no duty to anyone to
use the defective brakes, and did not violate any duty to anyone in failing
to use them." '' These assertions are
not plausible. In light of the obvi-
ous risks, a court would almost certainly decide that X owed a duty to P
and to everyone else on the road not to speed, and that X also owed a
duty to P to try to use the brakes to avoid hitting P. A jury would

107. E.g., Crowe, supra note 105; Green, supra note 8; Thode, supra note 100; see also L.
GREEN, W. PEDRICK, J. RAHL, E.W. THODE, C. HAWKINS, A. SMITH & J. TREECE, CASES ON THE
LAW OF TORTS (2d ed. 1977) (no discussion of actual causation issue).
108. Green, supra note 8, at 569 n.77.
109. Id.
110. Id.
111. Id.
1762 CALIFORNIA LAW REVIEW (Vol. 73:1735

undoubtedly conclude that X breached his duties to P by speeding and

failing to try to use his brakes. The implicit assumption in Green's state-
ment is that there is no duty and no negligence if performance of the
duty, in hindsight, would not have made any difference, even when the
defendant was unaware that performance would not make a difference.
But this is precisely the causal issue, which has been smuggled into the
duty analysis. 12 The duty and breach of duty are clear. Under the crim-
inal law, X would be liable for his speeding and failure to brake whether
or not they contributed to P's injury. The troublesome issue under the
tort law is the causal one: did X's speeding or failure to try to use his
brakes (the negligent aspects of his conduct) contribute to P's injury,
even though the injury would have occurred anyway given the defective
brakes?113 This causal issue is hardly clarified by the use of duty lan-
guage and duty analysis.
The confusion introduced by trying to resolve causation problems
through duty-risk analysis is further exemplified by Green's analysis of
D's tort liability. He notes initially that "D was under a duty to P and to
X and other users of the highway to refrain from selling the car to X
knowing it had defective brakes without disclosing that fact to X." But,
he asks,
Was the risk of driving at a speed which would not permit X to avoid
collision with P's car within the scope of D's duty to X and to P... ?
Would X have bought and driven the car if D had disclosed to him that
the brakes were defective? Under the evidence this could be a jury
issue. . . . [I]t cannot be presumed that he would have risked his own
safety and that of other users of the highway by driving a car with defec-
tive brakes. The risk of driving the car on the highway, irrespective of
the speed with which it was driven, falls within the scope1 of
14
D's duty to
X and to P with respect to the injuries suffered by them.
Green apparently concludes that D is liable for any injuries inflicted
on others by X's driving, no matter how negligently X drove, because
presumably X would not have been driving if D had told him about the
defective brakes. This again is a causal ("but for") argument masquerad-
ing as a duty issue. Moreover, it is a very weak causal argument.
Assuming that X would not have bought and used the particular car if he
had been told about the defective brakes, he presumably would have been
driving another car: his old car, a different new car, or a leased or bor-
rowed car. There is no reason to believe that the particular car had any-
thing to do with X's speeding or failure to brake. Thus, although D's

112. For another unsuccessful attempt to replace causal analysis with tortious-conduct analysis,
which also relies on causal arguments, see Landes & Posner, supra note 9, at 110- 11, 115-16, 119-22,
criticized in Wright, supra note 9, at 452-55.
113. The answer is yes. See infra text accompanying notes 274-76.
114. Green, supra note 8, at 569 n.77.
1985] CAUSATION IN TORT LAW 1763

failure to disclose was negligent, it was not a cause of P's injury.' 15

The above discussion reflects Green's final views on causation.
These views, published in 1962, had evolved over several decades as he
led the legal realists' fight to unburden causal analysis of all the critical
policy issues. In his initial writings, he clearly used the negligent aspect
of the defendant's conduct, rather than the defendant's conduct as a
whole, as the starting point for the cause-in-fact inquiry."I6 Even in his
1962 article, he continued to focus on the negligent aspect in the cases
involving failure to provide fire escapes, life saving equipment, lighting
for stairways, and so forth. In each of these cases, the failure to provide
the safety precaution is the negligent aspect of the defendant's overall
17
activity of operating a hotel, boat, store, train station, or whatever."
These cases, and the example discussed above, demonstrate the impor-
tance of focusing the causal inquiry on the tortious aspect of the defend-
ant's conduct. Failure to do so burdens the duty analysis with causation
issues that may well be mishandled since they are not explicitly recog-
nized as causation issues.

2. Keeton's "'HarmWithin the Risk" Theory

Similar problems appear in Robert Keeton's version of the risk the-
ory of liability. Keeton employs risk analysis in the proximate-cause
inquiry rather than in the duty-negligence (tortious-conduct) inquiry.
Initially, he recognizes the courts' insistence that the plaintiff's injury has
been caused by the negligent aspect of the defendant's conduct. The first
two formulations of his Risk Rule appear to be successively more explicit
statements of this requirement." 8 However, he eviscerates the require-
ment by interpreting "negligent aspect" to mean the type of risk that
115. See Saunders Sys. Birmingham Co. v. Adams, 217 Ala. 621, 117 So. 72 (1928). For an
analysis of this example under the causal test proposed in this Article, see infra text accompanying
notes 274-76.
116. L. GREEN, JUDGE AND JURY 188, 229-30 (1930); L. GREEN, supra note 2, at 3-4, 41, 145-
46, 148-49, 151-52, 164, 166, 181-84 n.60.
117. Green, supra note 8,at 559. But see id. at 560 n.45 (citing Stacy v. Knickerbocker Ice Co.,
84 Wis. 614, 54 N.W. 1091 (1893) (failure to provide warning fence)). Wayne Thode has applied
Green's later views to the cases involving failure to provide a fire escape. He asserts that furnishing
the room to the deceased is sufficient to satisfy the causal requirement, even when the failure to
provide the fire escape did not contribute to the victim's death. Thode, supra note 100, at 433-34.
Thode tries to convince those who disagree with this approach by arguing that we would surely hold
the defendant liable if the deceased stayed in his room because he knew there was no fire escape, yet
"D's conduct has not changed." D's conduct has not changed, but its causal effects have. The
failure to provide the fire escape has provided the reason for the deceased to stay in his room-a
causal connection that is lacking when it is assumed that the deceased stayed in his room for other
reasons. See infra note 212.
118. The first formulation states: "A negligent actor is legally responsible for that harm, and
only that harm, of which his negligence is a cause in fact." R. KEETON, supra note 104, at 4
(emphasis in original). The second formulation replaces "his negligence" with "the negligent aspect
of his conduct." Id. at 9 (emphasis in original).
1764 CALIFORNIA LAW REVIEW [Vol. 73:1735

made the defendant's conduct negligent, rather than the aspects of the
conduct itself that created that type of risk. 9 He then restates the
requirement in a third formulation that is very similar to Green's
approach:
A negligent actor is legally responsible for the harm, and only the harm,
that not only (1) is caused in fact by his conduct [as a whole] but also (2)
is a result within the scope
120
of the risks by reason of which the actor is
found to be negligent.
Under this formulation of the Risk Rule, the only causal require-
ment is that the injury have been caused by the defendant's conduct as a
whole. The second part of the formula merely requires that the injury fit
the description of one of the risks that made the defendant's conduct
negligent. It calls for a comparison of the injury with the relevant risk
description (the so-called "negligent aspect" of the conduct), rather than
an inquiry into the causal connection between the injury and some prior
2
act, omission, or condition.' 1
For example, Keeton discusses a hypothetical case in which the
defendant negligently placed an unlabeled can of rat poison next to food

119. Thus, in his primary example, Keeton describes the negligent aspect as "placing the poison
where it was likely to be mistaken for something intended for human consumption." Id. at 5; see id.
at 8-9. This is a description in terms of the risk created, rather than a description of the conduct
(placing unlabeled poison near food) which created the risk. See also id. at 14-16, 126-27 n.l 1.
120. Id. at 10. For Keeton's distinction between the conduct as a whole and the negligent
aspect of the conduct, see id. at 4-5.
There are some important differences between Keeton's and Green's respective approaches.
For example, Keeton limits the relevant risks to those that were foreseeable at the time the defendant
acted (or failed to act) and were the basis for characterizing the defendant's conduct as negligent.
That is, his "harm within the risk" proximate-cause analysis builds on the foreseeability analysis
employed in the determination of negligence. Id. at 10-11, 18-22, 51, 56. Green, on the other hand,
would identify the relevant risks as part of a broad social-policy analysis of the duty issue, in which
foreseeability of the risk is not necessary and is only one of many considerations. Foreseeability of
the type of consequence, but not the manner of its occurrence, is an important part of Green's
elaboration of the negligence (breach of duty) issue. The negligence-foreseeability formula, however,
is seen merely as a useful method for presenting the ultimate issue of culpability or responsibility to
the jury, which can and often does impose liability despite lack of foreseeability of the injury. Green,
supra note 8, at 562-64, 566-68 & n.72, 570-74; Green, Foreseeabilityin Negligence Law, 61 COLUM.
L. REV. 1401 (1961).
121. R. KEETON, supra note 104, at 48; see id. at 52-59. This confusion of causal analysis with
"harm within the risk" analysis leads Keeton to assert that "the Risk Rule is indeed a rule of
causation in a cause-in-fact sense" and that the proximate-cause inquiry is therefore a causal inquiry.
Id. at 13; accord, Robinson, supranote 7, at 756 n. 143, 757 n. 144; see R. KEETON, supra note 104, at
17-18, 81. Unfortunately, this confusion now pervades the fifth edition of Prosser's hornbook, of
which Keeton is a coeditor. Prosser's longstanding insistence that the actual-cause inquiry is factual
and that the proximate-cause inquiry (as distinct from the preliminary issue of actual causation) is
noncausal and policy-dependent has been systematically replaced by statements regarding the policy-
dependent nature of the actual-cause inquiry and the causal nature of at least part of the proximate-
cause inquiry. Compare W. PROSSER, supra note 1, § 41, at 237; id. § 42, at 244, 249-50; id. § 43, at
250-51; id § 45, at 289, with PROSSER & KEETON, supra note 7, at xix; id. § 41, at 264-65; id. § 42,
at 273, 274, 279-80; id. § 43, at 280-81; id. § 45, at 321.
1985] CAUSATION IN TORT LAW 1765

on a shelf in the kitchen of his restaurant. The shelf was next to a hot
stove, and the heat from the stove caused the can of poison to explode.
The force of the explosion killed a delivery man. The defendant did not
know, arid had no reason to know, that the rat poison might explode.
The defendant's conduct as a whole-placing the can on the shelf or,
broader yet, operating the restaurant-was a cause of the injury. How-
ever, Keeton states, the defendant is not liable because the description of
the injury ("injury by explosion") does not match the description of the
risk that made the defendant's conduct negligent ("injury by
poisoning"). 122
But suppose a customer died by poisoning as a result of eating the
rat poison, which was deliberately served to him by someone who knew it
was poisonous. The defendant's negligence did not contribute to the
injury, yet the defendant will be liable under Keeton's Risk Rule unless
the risk is described more precisely, for example, "injury by accidental
poisoning." Suppose then that a customer is poisoned accidentally and
nonnegligently as a result of being fed some substance other than the rat
poison. The risk must be detailed even further: "injury by accidental
poisoning with the rat poison that the defendant placed on the shelf."
What if a government health inspector accidentally contaminates the
customer's food with the rat poison, after opening and inspecting the can,
which had been properly labeled and moved away from the food several
days earlier? We must add a qualifier, "as a result of mistaking the
poison for food," to the previous description of the risk. Suppose the
government inspector mistook the poison for food even though it was
now away from food and properly labeled. We must add a further quali-
fier: "because the rat poison was unlabeled or near food."
At this point, we have included in the description of the risk a
description of the aspect of the defendant's conduct which made that
conduct negligent: having unlabeled poison around, or having poison
(even if properly labeled) near food. Moreover, we have included in the
description the requirement that the negligent aspect have been a cause
of the victim's injury: "injury. . . because the rat poison was unlabeled
or near food." Unless the description of the risk explicitly or implicitly
includes this tortious-aspect causation requirement, the Risk Rule will
encounter difficulties similar to those described in the previous para-
graph. Contrary to the courts' practice, the rule will permit the defend-
ant to be held liable even though the tortious aspect of his conduct was
not a cause of the injury. The same problem exists with respect to the
first and second formulations of Keeton's Risk Rule, since Keeton inter-
prets the terms "negligence" and "negligent aspect" in them to mean the

122. R. KEETON, supra note 104, at 3, 10.

1766 CALIFORNIA LAW REVIEW [Vol. 73:1735

type of risk that made the defendant's conduct negligent, rather than the
23
aspects of the defendant's conduct which created that type of risk.'
Keeton occasionally incorporates the tortious-aspect causation
requirement into the description of the risk, in order to avoid liability
that otherwise would exist under the Risk Rule. 124 More often, his dis-
cussions of specific cases implicitly assume that the injury not only must
have been within the broadly stated risk, but also must have been caused
25
by the tortious aspect of the defendant's conduct or activity.'
In sum, the Risk Rule mandates results that are inconsistent with
those reached by the courts, unless the tortious-aspect causation require-
ment is incorporated into the description of the risk. But, as with
Green's duty analysis, this awkward and circuitous method of posing the
causal inquiry creates a substantial danger of misperceiving or mishan-
dling the causal element in the liability analysis. Instead, as some propo-
nents of the risk theory have recognized, 126 the "harm within the risk"
limitation should be viewed as a (proximate-cause) supplement to the
tortious-aspect causation requirement, rather than as a substitute for it.

3. The Tortious-Aspect Causation Requirement

As far as I am aware, only two of the many books and articles on

causation in tort law have emphasized the existence and significance of

123. See supra note 119.

124. For example, Keeton discusses the case in which the victim negligently sat on an unsafe
(weak) wall, but was injured when the defendant crashed into the wall with a force that would have
knocked the wall down even if it had been safe. He describes the victim's negligence, in risk
language, as "placing himself where he was likely to be injured by the collapse of the wall, either
without an external impact or under an external impact insufficient to cause the collapse of a safe
wall," and he notes that "[tihis aspect of his conduct was not a sine qua non of the injury he
suffered." R. KEETON, supra note 104, at 127 n.11.
Of course, the victim's placing himself on the wall was a sine qua non of the injury. What
Keeton means is that the risk which made his conduct negligent did not materialize. To reach this
result, Keeton must define the risk more specifically than "injury due to collapse of the wall." His
definition must include the requirement that the collapse be caused by the unsafe condition of the
wall: the tortious-aspect causation requirement. (The description of the tortious aspect of the
defendant's conduct must include the unsafe condition of the wall, but the unsafe condition was not
a sine qua non of the injury. Nevertheless, it may have contributed to the injury. See infra text
accompanying notes 268-73.)
125. See R. KEETON, supra note 104, at 6, 10, 49-78.
126. E.g., Williams, supra note 21, at 62-65, 70-71, 73-75, 79; Williams, The Risk Principle, 77
LAW Q. REV. 179, 179 n.4 (1961); see also Malone, supra note 27, at 370-74. For a spirited
exchange between a proximate-cause risk theorist and a duty-negligence risk theorist, see
Henderson, supra note 105, and Thode, A Reply to the Defense of the Use of the HypotheticalCase to
Resolve the CausationIssue, 47 TEx. L. REv. 1344 (1969). While both Henderson and Thode insist
that the cause-in-fact inquiry is concerned only with the causal effect of the actor's conduct as a
whole, they each correctly accuse the other of relying heavily on the tortious-aspect causation
requirement in his risk analysis, and they note the unnatural and awkward language that the other is
required to use when the causal inquiry is incorporated into the risk analysis.
19851 CAUSATION IN TORT LAW

the tortious-aspect causation requirement, and in each instance the tor-

tious-aspect analysis is applied only to negligence cases.
The first of these is Charles Carpenter's extremely illuminating arti-
cle,' 27 in which he attempts, "without imposing a preconceived theory, to
restate, in consistent and workable form the principles [of actual cause
and proximate cause] . . . actually found in the decisions."12 Although
in a few places he confuses the "harm within the risk" issue with the
negligent-aspect causation requirement, 2 9 he subsequently clearly distin-
guishes the two issues and demonstrates the pervasiveness and greater
significance of the negligent-aspect causation requirement. 130 However,
he does not explain how one identifies the negligent aspect of conduct.
The second is Arno Becht and Frank Miller's book on actual-causa-
tion analysis in tort law. Becht and Miller suggest that the negligent
aspect of a person's conduct should be identified as the minimum aspect 3
which would need to be changed to make the conduct nonnegligent.1 1
The usefulness of this approach in clarifying the liability analysis is
demonstrated throughout their book.1 32 However, their conception of
the negligent aspect of conduct is too broad. They argue that the negli-
gent aspect is always an unqualified act or omission. They do not further
specify the negligent aspect by including the circumstances or conditions
which have to be combined with the act or omission in order to charac-
terize the conduct as negligent.
For example, Becht and Miller discuss a case in which the plaintiff
negligently sat on an unsafe (weak) wall, but was injured when the
defendant knocked down the wall with his car. They insist that the neg-
ligent aspect of the plaintiff's conduct was the act of sitting on the wall.

127. Carpenter, Workable Rules for DeterminingProximate Cause (pts. 1-3), 20 CALIF. L. REV.
229, 396, 471 (1932).
128. Id. at 257-58.
129. Id. at 231 (loaded gun discussion); id. at 231 n.5 (Butz case); id. at 253-54 (Teis case).
130. Id. at 408-19, 471-539.
131. A. BECHT & F. MILLER, supra note 8, at 34; see also id. at 12-13, 27-28 & n.40, 87-90.
132. E.g., id.at 34-42, 54, 58, 59-61, 85, 90-91, 140, 169-86. The authors' discussion of Kernan
v. American Dredging Co., 355 U.S. 426 (1958), is particularly interesting. The defendant's act of
carrying a kerosene signal lantern at a height lower than that mandated by statute was deemed
sufficient to hold him liable for the ignition of petroleum vapors on the surface of the water, even
though the purpose of the statute was to prevent collisions between ships, not the ignition of surface
vapors. Becht and Miller, expanding upon a footnote by Justice Harlan, 355 U.S. at 442 n.1, note
that the negligence was the omission of a signal lantern at the required height, not the presence of
one at the lower height (which was not in itself prohibited). If there had been a lantern at the
required height in addition to the one at the lower height, there would have been no negligence
(breach of the statute), yet the vapors would still have been ignited. Thus, the negligent aspect was
not a cause of the injury. A. BECHT & F. MILLER, supra note 8, at 38-41. But see H.L.A. HART &
T. HONORt, supra note 8, at 289 (failing to note this point). If the defendant's attorney had
correctly identified the negligent aspect and stressed the lack of causal connection between the
negligence and the injury, rather than conceding negligent causation and relying on the "harm
outside the risk" argument, he probably would have won the case.
1768 CALIFORNIA LAW REVIEW [Vol. 73:1735

This clearly was a necessary condition for the plaintiff's injury. They
therefore conclude that the negligent-aspect causation requirement is sat-
isfied. They refuse to consider, as part of the causal inquiry, whether the
unsafe condition of the wall contributed to the injury, since the condition
of the wall was not a part of the plaintiff's conduct. 133
But the unsafe condition is a necessary element in the description of
the plaintiff's negligent conduct. The act of "sitting on a wall" is not in
itself negligent. The act of "sitting on an unsafe (weak, structurally
unsound) wall" is negligent, when it is coupled with the mental element
required for negligence (actual or constructive knowledge of the unsafe
condition). The act of sitting and the unsafe condition are both necessary
to make the plaintiff's conduct negligent. Therefore it must be estab-
lished that both the act and the condition contributed to the injury before34
it can be said that the plaintiff's negligence contributed to the injury.,
In general, the tortious aspect of a person's conduct or activity is a
cause of an injury only if each of its necessary elements (act, omission,
condition, or circumstance) contributed to the occurrence of the
injury. 135 If a certain element did not contribute to the injury, but was
necessary to make the conduct or activity tortious, then it cannot be said
that the tortious aspect of the conduct or activity was a cause of the
136
injury.
The actor's mental state (intent, or actual or constructive knowledge
of a certain risk), which is required in addition to the conduct or activity
itself in order for the conduct or activity to be considered tortious, does
not enter directly into the causal inquiry. Instead, the intent or knowl-
edge is used to determine which acts, omissions, conditions, and circum-
stances constitute the tortious aspect of the conduct or activity. For
example, in the explosive rat poison case, the actor was aware (or should
have been aware) of its poisonous nature, but was not aware (and had no
reason to be aware) of its explosive nature. Therefore, the poisonous
nature of the substance was a constituent of the tortious aspect of the
actor's conduct, but its explosive nature was not.
Becht and Miller's overly broad conception of the tortious aspect of
conduct forces them erroneously to concede negligent causation and to
133. A. BECHT & F. MILLER, supra note 8, at 183-84; see also J. FLEMING, supra note 4, at 192
& n.43; Malone, supra note 27, at 371.
134. H.L.A. HART & T. HONORt, supra note 8, at 210; see infra text accompanying notes 268-
73.
135. When there is more than one tortious aspect, each must be considered, and the tortious-
aspect causation requirement is satisfied if any of them contributed. For example, in the rat poison
case, there were two tortious aspects: (1) having unlabeled poison around and (2) having poison
(even if labeled) near food. If either tortious aspect contributed to the injury, the tortious-aspect
causation requirement is satisfied.
136. See H.L.A. HART & T. HONORI, supra note 8, at 117-20, 208-11; Carpenter, supra note
127, at 409-10.
1985] CAUSATION IN TORT LAW 1769

rely instead on the "harm outside the risk" rationale to limit liability in a
number of cases. 37 It also leads them to adopt Green's and Keeton's
approach to causation of injury in the strict liability cases. They assert
that, under strict liability, the causal inquiry is applied to the defendant's
conduct as a whole.138 Presumably, they again would use the "harm
outside the risk" argument to limit liability.
Properly understood, the tortious-aspect causation requirement
applies to all tort actions, not just to negligence actions.
In the traditional strict liability cases, the tortious aspect of an activ-
ity includes all the acts, omissions, and conditions necessary to make it
ultrahazardous or abnormally dangerous: for example, the keeping or
use of a highly explosive substance, a poisonous fugitive substance, a
highly radioactive or toxic substance, a vicious or wild animal, a foraging
animal, or a large bulk accumulation of liquid. That tortious aspect
must be a cause of the injury, as the original Restatement of Torts made
clear: "[O]ne who carries on an ultrahazardous activity is liable to
another. . . for harm resulting thereto from that which makes the activ-
' 139
ity ultrahazardous."
Unfortunately, however, the academic proponents of the risk theory
succeeded in rewording section 519 in the Restatement (Second) to make
it conform to Keeton's Risk Rule: "(1) One who carries on an abnor-
mally dangerous activity is subject to liability for harm to [another]...
resultingfrom the activity . . . .(2) This strict liability is limited to the
kind of harm, the possibility of which makes the activity abnormally dan-
gerous. '""4° This Risk Rule formula will produce unintended results

137. A. BECHT & F. MILLER, supra note 8, at 61-62 (plaintiff negligently stood on unrailed
portion of icy platform, but was injured by wall falling on that portion of platform rather than by
slipping on ice); id. at 150-51, 199 (accident occurs while plaintiff or defendant is engaged in secular
activity in violation of Sunday blue law). See also their confusing discussion of the license cases at
id. at 141-50, 194-99.
138. Id. at 46, 48, 168.
139. RESTATEMENT OF TORTS § 519 (1938) (emphasis added).
140. RESTATEMENT (SECOND) OF TORTS § 519 (1977) (emphasis added). Among the Advisers
to the Reporter for the Restatement (Second) were such prominent academic proponents of the risk
theory as Eldredge, James, Robert Keeton, Page Keeton, Malone, and Seavey.
The Restatement (Second) also dropped the language in the original Restatement which limited
liability to those whom the actor "should recognize as likely to be harmed" by the miscarriage of the
activity. However, the drafters apparently assumed that this limitation was retained by the "kind of
harm" language in § 519(2). Thus, comment e to § 519(2) refers to risk and injury "to those in the
vicinity" of a possible explosion, and the accompanying illustration 1 rejects liability when the actor
"has no reason to know of the presence of B's mink ranch nearby," even though "[t]he noise of the
blasting frightens the mink and the fright causes them to kill their young."
The mink illustration is based on Madsen v. East Jordan Irrigation Co., 101 Utah 552, 125 P.2d
794 (1942). The result seems inconsistent with RESTATEMENT (SECOND) OF TORTS § 522(b)
(liability despite unforeseeable contributing action of animal) and also the Risk Rule itself.
Frightened or startled reactions to the noise of blasting, as well as direct effects of the force of the
blast, are foreseeable and "within the risk," and the peculiar reaction of the mink in such situations
1770 CALIFORNIA LAW REVIEW [Vol. 73:1735

unless the tortious-aspect causation requirement is read back into it or

into the definition of each particular risk ("kind of harm"). 14 ' For exam-
ple, suppose a truck transporting dynamite nonnegligently runs over a
detonation device set up by someone else, thereby causing explosives
belonging to that other person, but not those in the truck, to go off and
cause injury in the vicinity of the truck. 142 Unless the tortious-aspect
causation requirement is read in, the owner of the truck will be strictly
liable under section 519. Once again, the "harm within the risk" princi-
ple can be at best a supplement to, rather than a substitute for, the tor-
tious-aspect causation requirement.
In strict product liability cases, the tortious aspect of the defendant's
conduct or activity is the defect in the particular product, and it is clear
that the defect must be a cause of the injury.' 4 3
In the intentional tort cases, the tortious aspect is all the acts, omis-
sions, conditions, and circumstances necessary to create the substantial
certainty of invasion of a protected interest, or the acts or omissions done
for the purpose of causing such an invasion. The substantial-certainty
variety of intentional tort is analogous to negligence, with one difference.
The risk created in the intentional tort action must be substantially cer-
tain to be realized but need not be unreasonable, while the risk created in
the negligence action can be much less likely but must be
unreasonable.'44
Finally, in the nuisance cases, liability is based on conduct that is
either negligent, abnormally dangerous, or intentional (usually in the
sense of substantially certain although reasonable).' 4 5 Therefore, the tor-

is a well-established characteristic. See, e.g., Wildwood Mink Ranch v. United States, 218 F. Supp.
67 (D. Minn. 1963); Gronn v. Rogers Constr., 221 Or. 226, 350 P.2d 1086 (1960); Summit View,
Inc. v. W.W.Clyde & Co., 17 Utah 2d 26, 403 P.2d 919 (1965); Foster v. Preston Mill Co., 44 Wash.
2d 440, 268 P.2d 645 (1954); MacGibbon v. Robinson, [1952] 4 D.L.R. 142 (B.C.); R. KEETON,
supra note 104, at 135 n. 100. In the Madsen case, the defendant irrigation company presumably was
familiar with the area and, at any rate, was blasting only 100 yards from the mink ranch. Both the
vibrations and the noise terrified the mother mink, which reacted in a natural way (for mink). The
result is justified only as an application of the principle that a person need not take excessive care to
avoid injury to extrasensitive plaintiffs and, correspondingly, should not be held strictly liable for
"harm [that] would not have resulted but for the abnormally sensitive character of the plaintiffs
activity." RESTATEMENT (SECOND) OF TORTS § 524A (1977).
141. See supra text accompanying notes 122-23.
142. Cf RESTATEMENT (SECOND) OF TORTS § 519 comment e (1977); R. KEETON, supra note
104, at 105.
143. Greenman v. Yuba Power Prods., 59 Cal. 2d 57, 62-64, 377 P.2d 897, 900-01, 27 Cal. Rptr.
697, 700-01 (1962); Codling v. Paglia, 32 N.Y.2d 330, 342, 298 N.E.2d 622, 628-29, 345 N.Y.S.2d
461, 469-70 (1973); RESTATEMENT (SECOND) OF TORTS § 402A (1965).
144. Thus, it is not true, as Keeton claims, that "[n]o concept of risk is resorted to in
determining the issue of liability for intentional tort." R. KEETON, supra note 104, at 100-01.
145. Boomer v. Atlantic Cement Co., 26 N.Y.2d 219, 257 N.E.2d 870, 309 N.Y.S.2d 312
(1970); Morgan v. High Penn Oil Co., 238 N.C. 185, 77 S.E.2d 682 (1953); Jost v. Dairyland Power
1985] CAUSATION IN TORT LAW 1771

tious aspect is defined according to the type of tortious conduct involved

ineach case.

4. The Power of the Tortious-Aspect CausationRequirement

The tortious-aspect causation requirement resolves many of the
cases that have been prominent in discussions of the proximate-cause
limits on tort liability, especially in the writings of the risk theorists.
There is thus no need to resort to noncausal policies or principles under
the rubrics of proximate cause, duty, recoverable damages, and so forth
in order to deny liability in these cases. In each of the following exam-
ples, the italicized condition was a necessary constituent of the tortious
aspect of the defendant's or the plaintiff's conduct, but the condition did
not contribute to the injury, so the tortious aspect was not a cause of the
injury. The defendant handed a loaded gun to a child, who dropped the
gun on his foot.14 6 The defendant's box of explosives fell on the plaintiff's
foot, or was being transported in a truck that ran into the plaintiff (no
47
explosion occurred). The plaintiff stood on the icy, unrailedportion of
a platform, where a wall fell on him."' The plaintiff stood in an area
where trucks sped through, but was hit by a railroad freight car that came
through the wall. 14 9 The defendant tested an unguarded fan on the floor
where men with loose clothing had to bend over it (creating a risk that the
clothing might get caught in the fan), but the plaintiff stuck his fingers
150
into the fan.
Coop., 45 Wis. 2d 164, 172 N.W.2d 647 (1969); RESTATEMENT (SECOND) OF TORTS §§ 822, 825,
826 (1977); see also Renken v. Harvey Aluminum, 226 F. Supp. 169 (D. Or. 1963).
146. RESTATEMENT (SECOND) OF TORTS § 281 illustration 3 (1965); L. ELDREDGE, MODERN
TORT PROBLEMS 18 (1941); J. FLEMING, supra note 4, at 192; 2 F. HARPER & F. JAMES, supra note
2, § 20.5, at 1136-37; Carpenter, supra note 127, at 231.
147. RESTATEMENT (SECOND) oF TORTS § 519 comment e (1977); Seavey, Mr. Justice Cardozo
and the Law of Torts, 39 COLUM. L. REv. 20, 35, 52 HARV. L. REV. 372, 387, 48 YALE L.J. 390,
405 (1939); Williams, supra note 21, at 70-71.
148. Smithwick v. Hall & Upson Co., 59 Conn. 261, 21 A. 924 (1890), discussed in A. BECHT &
F. MILLER, supra note 8, at 61-63; L. ELDREDGE, supra note 146, at 21-22; H.L.A. HART & T.
HoNORf,supra note 8,at 209.
149. Gray v. Scott, 66 Pa. 345 (1870), discussed in L. ELDREDGE, supranote 146, at 22. Almost
all the cases discussed by Eldredge in his argument on behalf of the risk theory involve a lack of
causation by the tortious aspect of the actor's conduct. In addition to the Gray case, see the
examples cited supra notes 146 & 148 and Falk v. Finkelman, 268 Mass. 524, 168 N.E. 89 (1929)
(car parked overtime was knocked into another car by fire engine); New York, L.E. & W.R. Co. v.
Ball, 53 N.J.L. 283, 21 A. 1052 (1891) (passenger in baggage compartment in violation of rule was
injured in train wreck; injury unaffected by being near baggage); Hudson v. Lehigh Valley R.R., 54
Pa. Super. 107 (1913) (bicyclist who failed to "stop, look and listen" for train was hit by gate
negligently lowered onto him). These cases are all discussed in a section entitled "The Hazard
Problem," in L. ELDREDGE, supra, at 17-24, which also is the title of RESTATEMENT (SECOND) OF
TORTS § 281 comment e, for which Eldredge was an Adviser. See supra note 140.
150. Thurogood v. Van den Berghs & Jurgens Ltd., [1951] 2 K.B. 537 (C.A.), discussed in
Goodhart, The Imaginary Necktie and the Rule in Re Polemis, 68 LAW. Q. REV. 514 (1952);
Williams, supra note 21, at 71-72. Goodhart and Williams assume that the loose-clothing element
1772 CALIFORNIA LAW REVIEW [Vol. 73:1735

The tortious-aspect causation requirement similarly resolves many

of the troublesome statutory-violation (negligence per se) cases. Initially,
it should be noted that many statutes are not passed for safety pur-
poses1 51 or are not meant to lay down rigid rules of behavior, but rather
are meant to be administered flexibly. In such cases, there is no basis for
inferring negligence from violation of the statute, so the issues of actual
cause and proximate cause never arise. Neither the tortious-aspect cau-
sation requirement nor the "harm within the risk" limitation is relevant
or needed to deny liability. However, even if the violation is erroneously
treated as tortious conduct, the tortious-aspect causation requirement
rarely will be satisfied. For example, if an injury occurred while the
defendant or the plaintiff was engaged in an activity on Sunday (in viola-
tion of a Sunday blue law), had parked his car at a meter without paying,
or was hunting without a permit, the italicized element in each instance
was a necessary constituent of the statutory violation but did not contrib-
ute to the injury. Therefore, the "tortious" aspect of the conduct was not
a cause of the injury.15 The few cases which nevertheless impose liabil-
ity on the defendant or refuse to allow recovery by the plaintiff because of
the statutory violation ignore tort liability principles in order to further
the purposes of the criminal law. These cases, however, generally have
been repudiated." 3
The failure to satisfy the tortious-aspect causation requirement also
explains the denial of liability in many cases where the statute was passed
for safety purposes, but a necessary element of the illegal or tortious
aspect of the defendant's conduct did not contribute to the injury. For
example, Prosser cites cases in which the defendant set out a poisonous
substance, which exploded; or failed to erect a fence next to a railroad
track and a cow died from eating too much; or sold explosive materials
(fireworks), which a child ate. 5 4

was necessary to the finding of negligence, but this is not clear. If "unguarded operating fan on floor
near workmen" was sufficient for negligence, the tortious-aspect causation requirement was satisfied
and liability seems appropriate, absent contributory negligence. See H.L.A. HART & T. HONORt,
supra note 8, at 118-19.
151. W. PROSSER, supra note 1, § 36, at 192-93, 196.
152. H.L.A. HART & T. HONORC, supra note 8, at 210-11, 293-94; Carpenter, supra note 127,
at 409-10, 414-15. Writers who assert that the illegal conduct was a cause of the injury erroneously
apply the causal inquiry to the conduct as a whole rather than to the illegal aspect of such conduct.
E.g., A. BECHT & F. MILLER, supra note 8, at 150-51, 199; L. GREEN, JUDGE AND JURY 234-35
(1930); W. PROSSER, supra note 1, § 36, at 193.
153. 2 F. HARPER & F. JAMES, supra note 2, § 17.6, at 995-97; H.L.A. HART & T. HONORt,
supra note 8, at 210-11; W. PROSSER, supra note 1, § 36, at 193 n.53, 202-03. But see G.
CALABRESI, supra note 100, at 267-74, 301-08; Cole, supra note 7, at 459-62, 598-607; Green, supra
note 8, at 569 n.77; Morris, supra note 3, at 1096-97; Thode, supra note 126, at 1348-49, 1351-52;
Thode, supra note 100, at 430-31. For a cogent critique of Thode's argument, see Strachan, supra
note 7, at 389-91.
154. W. PROSSER, supra note 1, § 36, at 195-96 & n.77.
1985] CAUSATION IN TORT LAW 1773

In the above cases, it can also be said that the harm was not "within
the risk." The "harm within the risk" theory, however, cannot account
for some other cases. Thus, if an unlicensed driver causes an automobile
accident or an unlicensed medical practitioner causes a medical mishap,
the harm is the sort which the legislature meant to prevent through its
licensing requirements. To describe the risk as restricted to "injury at
the hands of incompetents"' 5 5 is to rely on and incorporate the tortious-
aspect causation requirement. The tortious aspect is "driving or practic-
ing medicine without a license." Although the overall conduct of driving
or practicing medicine contributed to the injury, 5 6 the failure to have the
required piece of paper (the license) did not.157
A similar problem exists when the defendant is violating the statu-
tory speed limit, but a child darts in front of his car whom he could not
avoid even if he were driving at the legal speed. The injury in such cases,
as even Green admits, 5 ' is "within the risks" meant to be avoided by the
speed limit. Some argue that the injury is not within the risks, which are
said to be limited to those involving "lack of control."'5 9 But the risks
meant to be avoided include not only literal loss of physical control of the
car but also inability to swerve or stop in time to avoid hitting something
or someone. The injury is "within" this latter risk, but there is no liabil-
ity since the tortious aspect (excess speed) did not contribute to the
injury. Liability can be avoided under the "harm within the risk"
approach only by incorporating the tortious-aspect causation require-
ment in the description of the risk. 6 °
When the tortious-aspect causation requirement is satisfied in a par-
ticular case, the courts often hold the defendant liable even though the
injury was unforeseeable or "outside the risk," especially when the tor-

155. Id. at 196; PROSSER & KEETON, supranote 7, § 36, at 226; see Green, supranote 8, at 547-
48.
156. PROSSER & KEETON, supra note 7, § 36, at 223-24; Williams, supra note 21, at 74-75; see
A. BECHT & F. MILLER, supra note 8, at 141-42.
157. H.L.A. HART & T. HONORt, supra note 8, at 117-20; Carpenter, supra note 127, at 412-13,
414. A legislature or court might decide, however, that the lack of license should be treated as some
evidence or even as prima facie evidence that the driver or practitioner was incompetent to engage in
the activity and that the injury was caused by some unspecified instance of that incompetence
(negligence). E.g., N.Y. CIv. PRAC. LAW § 4504(d) (McKinney Supp. 1984); 2 F. HARPER & F.
JAMES, supra note 2, § 20.3, at 1120-21; Williams, supra note 21, at 74-75. But see A. BECHT & F.
MILLER, supra note 8, at 143-50.
158. Green, Are There Dependable Rules of Causation? 77 U. PA. L. REV. 601, 619 (1929).
159. E.g., 2 F. HARPER & F. JAMES, supra note 2, § 20.5, at 1138 & n.16, 1148; id. § 20.6, at
1156 n.23.
160. See Cole, supra note 7, at 766-67; Henderson, supra note 105, at 189-92, 195-97; Thode,
supra note 126, at 1345-47; Thode, supra note 100, at 427-28, 430-31. To avoid the argument that
the excess speed did contribute to the injury by bringing the defendant to the precise place where the
child darted out at the precise time that he darted out, I assume here that the defendant had just
begun speeding. Otherwise, the tortious-aspect causation requirement is satisfied, and the defendant
must rely on noncausal proximate-cause arguments to avoid liability.
1774 CALIFORNIA LAW REVIEW [Vol. 73:1735

tious conduct was intentional or negligent. 161 Even the risk-theory pro-
ponents acknowledge this.1 62 In the statutory context, this approach is
reflected in expansive interpretations of the statutory purposes, 163 or
occasionally even by frank imposition of liability for injuries conceded to
be "outside the risk."'" Thus, the causal inquiry, focused on the tor-
tious aspect of the defendant's conduct, plays an extremely significant
role in both establishing and limiting legal responsibility.

II
THE NATURE AND CONTENT OF THE CAUSAL INQUIRY
In Part I of this Article, I analyzed the role that the causal inquiry
occupies in tort-liability analysis. In this Part I examine the nature and
content of the causal inquiry itself. As will be seen, efforts to articulate a
comprehensive, factual test of actual causation have failed repeatedly.
The causal inquiry therefore increasingly has been viewed by academics
as just one more manipulable policy tool that has little or no inherent
165
substantive content.
In this Part I attempt to rebut this prevailing view by elaborating a
causal test that incorporates the traditional Humean philosophic account
of the meaning of causation, as modified by John Stuart Mill. This test,
which I call the "NESS" (Necessary Element of a Sufficient Set) test, was
first suggested by Hart and HonorS. As elaborated here, the NESS test
states that a particular condition was a cause of a specific consequence if
and only if it was a necessary element of a set of antecedent actual condi-
tions that was sufficient for the occurrence of the consequence. A sub-
stantial portion of this Part is devoted to developing this test and
applying it to resolve the problematic causation cases that have resisted
solution under all the previously proposed tests.
First, however, the previous tests are examined in order to provide a
basis for comparison with the NESS test, to introduce the various types
of problematic causation cases, and to develop some of the concepts and

161. E.g., In re Kinsman Transit Co., 338 F.2d 708, 723-25 (2d Cir. 1964), cert. denied, 380
U.S. 944 (1965); H.L.A. HART & T. HONORt, supra note 8, at 176-78, 254-90; W. PROSSER, supra
note 1, §§ 43-44, at 250-80; Smith, supra note 106, at 127-28, 233-52, 321-27. Some proximate-cause
limitations on liability do exist. The primary instances are when the injury would not have occurred
but for unforeseeable tortious conduct by others or independent abnormal events or conditions. See
H.L.A. HART & T. HONORA, supra note 8, passim; Carpenter, supra note 127, at 471-539.
162. See, eg., L. GREEN, supra note 2, at 177-85; 2 F. HARPER & F. JAMES, supra note 2,
§ 20.5, at 1139-51; id. § 20.6, at 1155-56, 1160-61; R. KEETON, supra note 104, at 28-32, 39, 49-54,
60-78, 96-97, 100-03, 109-10, 117; Green, Foreseeability in Negligence Law, 61 COLUM. L. REV.
1401, 1417-24 (1961); Williams, The Risk Principle,77 LAW Q. REV. 179, 181-87, 193-203 (1961).
163. 2 F. HARPER & F. JAMES, supra note 2, § 17.6, at 1004-05; PROSSER & KEETON, supra
note 7, § 36, at 226-27.
164. See the discussion of the Kernan case supra note 132.
165. See sources cited supra notes 6, 7 & 9.
1985] CAUSATION IN TORT LAW 1775

arguments that underlie the NESS test. The first section discusses the
traditional "but for" (necessary condition) test and distinguishes the two
types of overdetermined-causation cases that the test fails to handle prop-
erly: the duplicative-causation cases and the preemptive-causation cases.
Subsequent sections discuss modifications of or alternatives to the but-for
test.
Finally, in the last section, the steps involved in applying the NESS
test are analyzed in detail to demonstrate that, contrary to the currently
popular view, the hypothetical inquiry associated with the but-for neces-
sary-condition analysis does not undermine the factual nature of the
causal inquiry. The miscellaneous arguments advanced by Malone to
establish the policy-dependent nature of the causal inquiry are also
examined and rejected.

A. The Rut-For Test and Its Limits: Duplicative

and Preemptive Causation
The most widely used test of actual causation in tort adjudication is
the but-for test, which states that an act (omission, condition, etc.) was a
cause of an injury if and only if, but for the act, the injury would not have
occurred. That is, the act must have been a necessary condition for the
occurrence of the injury. The test reflects a deeply rooted belief that a
condition cannot be a cause of some event unless it is, in some sense,
necessary for the occurrence of the event. This view is shared by law-
yers, philosophers, scientists, and the general public.1 66
In the vast majority of cases, the but-for test works quite well as a
test of actual causation. But in certain types of cases, it results in a find-
ing of no causation even though it is clear that the act in question con-
tributed to the injury. These are cases of overdeterminedcausation: cases
in which a factor other than the specified act would have been sufficient
to produce the injury in the absence of the specified act, but its effects
either (1) were preempted by the more immediately operative effects of
the specified act or (2) combined with or duplicated those of the specified
act to jointly produce the injury.
I will refer to the first type of situation as a case of preemptive causa-
tion. For example, D shoots and kills P just as P was about to drink a
cup of tea that was poisoned by C. D's shot was a preemptive cause of
P's death; C's poisoning of the tea was not a cause because its potential
effects were preempted. I will refer to the second type of situation as a
case of duplicative causation. For example, C and D independently start
separate fires, each of which would have been sufficient to destroy P's
166. E.g., T. BEAUCHAMP & A. ROSENBERG, supra note 21, at 94-96, 98, 107-17, 119, 131-37,
139-45, 297; H.L.A. HART & T. HONORf, supranote 8, at 15-16, 69, 110; W. PROSSER, supra note 1,
at 237-39; Mackie, supra note 21, at 16-21, 25-27; Williams, supra note 21, at 63-64.
1776 CALIFORNIA LAW REVIEW [Vol. 73:1735

house. The fires converge and together burn down the house. Each fire
was a duplicative cause of the destruction of the house. Yet, application
of the but-for test would result in a finding that D's shot was not a cause
of P's death in the first example and that neither C's nor D's fire was a
cause of the destruction of P's house in the second example.
A different sort of objection to the but-for test focuses on the hypo-
thetical or counterfactual nature of the inquiry that the test calls for. A
number of writers have asserted that the hypothetical nature of the but-
for inquiry necessarily involves or at least invites introduction of policy
67
considerations into a supposedly factual inquiry.1
While it might be thought that these difficulties could be avoided by
substituting a sufficient condition test for the but-for necessary-condition
test, this substitution would eliminate almost every potential cause, since
few if any acts are sufficient by themselves to produce any particular con-
sequence. Thus, in the scientific and philosophic literature on causation,
it is usually stressed that the cause of an event must include all the condi-
tions which together are sufficient to produce the consequence. 168 Conse-
quently, both lawyers and philosophers often conclude that this scientific
or philosophic concept of causation is of little relevance or use in the
69
law. 1
On the other hand, the sufficient-condition test could be interpreted
to mean any condition that is sufficient in combination with other condi-
tions to produce the consequence, even though it is not sufficient by itself.
Under this interpretation, however, anything could be treated as a cause
1 70
simply by adding it to an already sufficient set of conditions.
Judges and legal writers have responded to the actual and perceived
deficiencies in the but-for test in several different ways. A few writers
have tried to improve the but-for test by modifying the manner in which
it is applied.' 71 Most judges and writers, however, have adopted the sub-
stantial-factor formula, either as a supplement to or as a substitute for
the but-for test depending on their degree of dissatisfaction with the but-
for test. 172 Others have relied on an undefined and irreducible notion of

167. E.g., J. FLEMING, supra note 4, at 172; Cole, supra note 7; Malone, supra note 6, at 67-68;
see also sources cited infra note 284.
168. E.g., T. BEAUCHAMP & A. ROSENBERG, supra note 21, at 85, 291; H.L.A. HART & T.
HONORt, supra note 8, at 17-18, 21 & n.16; W. PROSSER, supra note 1, § 41, at 237 n.8; Williams,
supra note 21, at 63, 66.
169. E.g., RESTATEMENT (SECOND) OF TORTS § 431 comment a (1965); T. BEAUCHIAMP & A.
ROSENBERG, supra note 21, at 283-94; A. BECHT & F. MILLER, supra note 8, at 9, 12; J. FLEMING,
supra note 4, at 171 & n.2, 192 & n.43; H.L.A. HART & T. HONOR~f, supra note 8, at 2, 9-16, 68-69;
Calabresi, supra note 9, at 69-70, 105-06; Epstein, supra note 13, at 160-61; Morris, supra note 3, at
1088; Williams, supra note 21, at 66-68, 75-76.
170. Sosa, Introduction, in CAUSATION AND CONDITIONALS 1-3 (E. Sosa ed. 1975).
171. See infra text accompanying notes 176-89.
172. See infra text accompanying notes 190-208.
1985] CA USA TION IN TORT LAW 1777

directly observable causal contribution. 173

Each of these approaches will be discussed in subsequent sections of
this Part. It will be demonstrated that, under each approach, the analy-
sis ultimately must fall back on the but-for test in order to be useful and
coherent. The necessary-condition element, in some form, indeed seems
to be fundamental to the concept of causation. But it obviously is too
restrictive when applied in the usual way to the overdetermined-causa-
tion cases. Some writers treat the overdetermined-causation cases as gen-
uine instances of causation that must be admitted as unexplained
exceptions to the but-for test.17 4 A larger and growing number of writers
treat the overdetermined-causation cases as policy-based exceptions to
the actual-causation requirement, which is equated with the but-for
test. 175

B. Efforts to Modify the But-For Test

As mentioned above, the but-for test works well as a test of actual
causation in the vast majority of cases, but produces obviously incorrect
results in the overdetermined-causation cases. A few'writers have tried
to extend the but-for test's range of application by modifying the way in
which it is applied. Rollin Perkins, Arno Becht, and Frank Miller apply
the test to a detailed description of the injury or the manner of its occur-
rence. Glanville Williams excludes consideration of hypothetical facts.
And the editors of the fifth edition of Prosser's hornbook on tort law
apply the test to multiple potential causes in the aggregate. None of
these approaches works. Several of them assume the very answer the test
is supposed to provide.

1. Detailing the Manner of Occurrence

Perkins tries to resolve the causal issue in the overdetermined-causa-
tion cases by detailing the manner in which the injury occurred: "When-
ever that would not have happened when and as it did happen, had it not

173. See infra text accompanying notes 209-25.

174. E.g., 2 F. HARPER & F. JAMES, supra note 2, § 20.2, at 1110; id. § 20.3, at 1122-23; W.
PROSSER, supra note 1, § 41, at 239-40; see also Carpenter, supranote 127, at 397-98, 402-03, 406-07.
175. E.g., C. MORRIS & C.R. MORRIS, supra note 4, at 187-90; PROSSER & KEETON, supranote
7, § 41, at 265 & n.8, 268-69; Calabresi, supranote 9, at 86-87; Delgado, supra note 7, at 886-87, 891-
92; Edgerton, supra note 3, at 373; Grady, supra note 9, at 804 & n.21; Malone, supra note 6, at 89-
97; McLaughlin, Proximate Cause, 39 HARV. L. REV. 149, 153-54 (1925); Peaslee, Multiple
Causationand Damage, 47 HARV. L. REV. 1127, 1130-32 (1934); Rosenberg, supra note 7, at 855 &
n.27, 863-64; Strachan, supra note 7, at 389-95; Williams, supra note 21, at 75-76. Some writers,
who rely on tests other than the but-for test, treat only the most difficult overdetermined-causation
cases as policy-based exceptions to the actual-causation requirement. A. BECHT & F. MILLER, supra
note 8,at 95-98, 210-11; J. FLEMING, supra note 4, at 174 n. 19; H.L.A. HART & T. HONORf, supra
note 8, at 7-8, 125, 239-41; Pedrick, supra note 7, at 645, 654, 658-59; Weinrib, supra note 7, at 529-
33.
1778 CALIFORNIA LAW REVIEW [Vol. 73:1735

been for this, this is an actual cause of that."176 For example, if a victim
is killed instantaneously by a bullet wound while already dying from a
knife wound, he would not have died, as he in fact did, at the earlier
moment and by a bullet wound if not for the firing of the bullet. Without
the bullet wound, he would have died later, from the effects of the knife
wound. This, however, did not in fact happen. Thus, the bullet wound,
but not the knife wound, is a but-for cause of the death "when and as" it
happened.
This form of reasoning is nothing more than proof by tautology.
For example, consider Perkins' treatment of the case in which the victim
is struck simultaneously by two bullets, each of which would have been
instantly fatal by itself: "[He] would not have died when and as he did
die (by two bullets) had only one been fired. 'One might have caused the
result, but in fact both did so.' "177 The factors believed to be causally
relevant (the two bullets) are incorporated into the description of the
manner of occurrence of the injury (death by two bullets), and they are
then demonstrated to be causally relevant because we cannot construct
that precise description without them.
Such an approach could just as easily have included the victim's
knife wound, his silk shirt, or the air temperature in the description of
the manner of occurrence of the injury. If so included, each of these
conditions would also be proven under Perkins' test to be but-for causes
of the harm. We do not include them because we do not believe they
were causally relevant. But that is precisely the issue the test is supposed
to determine. The test "proves" only what has already been decided. It
does not tell us how that decision was made or help us to make it in hard
78
cases. 1

2. Detailing the Injury

Becht and Miller accept Perkins' reasoning,' 7 9 but they focus more
on the details of the harm itself-the final result-than on the manner of
its occurrence. 8 ° For example, in the case of the two fires, one set by the
defendant and each sufficient to destroy the plaintiff's house, Becht and

176. R. PERKINS, CRIMINAL LAW 689 (2d ed. 1969) (emphasis in original).
177. Id. (emphasis in original) (citation omitted).
178. Hart and Honor6, although recognizing that prior causal determinations are hidden in this
method, seem to approve its use. H.L.A. HART & T. HONORt, supra note 8, at xli-xlii, 124-25, 235
& n.56, 252-53; accord J. FLEMING, supra note 4, at 173-74; R. KEETON, supra note 104, at 7; Rizzo,
The Imputation Theory of Proximate Cause: An Economic Framework, 15 GA. L. Rav. 1007, 1020
(1981).
179. A. BECHT & F. MILLER, supra note 8, at 17; see also id. at 210.
180. One of their examples, discussed in the text, refers to being hit by "different parts of the
car" (manner of occurrence) as well as to "slightly different injuries" (result), but almost all the
other examples refer to different details of the result. However, their suggested approach of using
the but-for test at short intervals throughout the causal process, rather than only once at the time the
1985] CAUSATION IN TORT LAW 1779

Miller assert that, using the but-for test applied "in minute detail, it
would probably appear that the defendant's fire was a cause, for the posi-
tions of the smoke, ashes, and some parts of the ruins might well have
been caused by [i.e., not have existed but for] the defendant's fire."18 1
Again, this assertion begs the question. Under this approach, the
defendant's fire was a but-for cause only of the precise "positions of the
smoke, ashes, and some parts of the ruins," not of the destruction of the
house, for which these precise details are irrelevant. The detailed
description will be useful only if the plaintiff wishes to recover for dam-
ages that would not have occurred but for the precise position of the
smoke and debris-a most unlikely lawsuit.
If being a but-for cause of even the most trivial detail of a harm were
sufficient to make the defendant a cause of all the harm, the detailed but-
for approach would prove too much. Becht and Miller discuss an exam-
ple in which an inattentive driver hits a pedestrian who suddenly ran into
the path of the driver's car. If the driver had been attentive, he could
have swerved a little, but not enough to avoid impact with serious injury.
Nevertheless, a detailed application of the but-for test shows that the
driver's inattentiveness "was a cause of the injuries actually suffered,"
since "the injuries would have been at least slightly different and would
have been inflicted by different parts of the car if the driver had
swerved."' 8 2 Becht and Miller admit that both laypersons and lawyers,
using the common knowledge of causation upon which Becht and Miller
rely so strongly, 183 would conclude that the driver's inattentiveness was
not a cause in such a case. They suggest that laypersons and lawyers
make this mistake by failing to pay careful attention to the precise
details.' 8 4 But laypersons and lawyers, even after considering the precise
details, would still assert correctly that the inattentiveness was not a
cause of the serious injury, although it may have been a (but-for) cause of
some details associated with the injury. These details have causal signifi-
cance only if they contributed to the seriousness of the injury.
Becht and Miller realize that they, rather than the laypersons and
lawyers, must give way on this issue. They compromise their usual
causal principles by "equating the injuries." They reluctantly disregard

process is completed, is similar to Perkins's detailing of the manner of occurrence. See id. at 15-17,
19.
181. Id. at 18.
182. Id. at 28. Becht and Miller use the phrases "hypothetical inquiry" and "construction of
parallel series," but they acknowledge that their approach is a narrow form of the but-for test. The
test is applied step-by-step and in minute detail, and it excludes consideration of causes that the actor
could not have prevented. Id. at 17-21, 32; see infra text accompanying notes 211 & 219.
183. A. BECHT & F. MILLER, supra note 8,at 9-11; see infra text accompanying notes 209-11,
217.
184. A. BECHT & F. MILLER, supra note 8,at 28-29.
1780 CALIFORNIA LAW REVIEW [Vol. 73:1735

minor differences and conclude that the inattentiveness was not a

cause." 5 Although they do not acknowledge it, they thereby destroy the
alleged usefulness of the minute-detail approach. After utilizing the min-
ute-detail approach to support a finding of causation, they then are
forced to ignore the details and reject causation by equating the injuries.
The concept of "equating the injuries" introduces an inconsistency into
their theory that undermines their use of the minute-detail approach to
support a finding of causation in the merged-fires case.

3. Excluding Hypothetical Facts

Glanville Williams also refers to Perkins' approach of detailing the
manner in which the harm occurred, but he proposes a different
approach for the preemptive-causation cases. In such cases, facts that
did not actually occur must not be "invented" and considered, no matter
how likely they were to occur. Thus, the hypothesized but almost certain
fact that P would have been shot by C if he hadn't already been shot by D
cannot be considered.18 6 However, on the very next page, Williams is
obliged to depart from this rule to reach the correct causal conclusion in
a case of negligent omission.'8 7 Moreover, such a rule is of no help at all
in cases where the preempted condition actually has occurred-for
example, when P actually was shot by C after he already had been shot
and killed by D. In this type of preemptive-causation case, and in the
duplicative-causation cases, Williams abandons the but-for test for actual
causation and treats the causal issue as a policy issue.' 8 8

4. Aggregating Multiple Potential Causes

The editors of the fifth edition of Prosser's hornbook on tort law
attempt to resolve the causal issue in the overdetermined-causation cases
by applying the but-for test to the aggregate of potential causes:
When the conduct of two or more actors is so related to an event that
their combined conduct, viewed as a whole, is a but-for cause of the
event, and application of the but-for rule to them individually would
absolve all of them, the conduct of each is a cause in fact of the event. 189
This version of the but-for test, like Perkins's version, works only if the
person applying the test already has correctly determined which factors

185. Id. at 29-31.

186. Williams, supra note 21, at 72.
187. Id. at 73.
188. Id. at 75-77.
189. PROSSER & KEETON, supra note 7, § 41, at 268; see also 2 F. HARPER & F. JAMES, supra
note 2, § 20.2, at 1110; Carpenter, supra note 127, at 406-07; Carpenter, Concurrent Causation, 83 U.
PA. L. REV. 941, 944-45, 952 (1935); cf A. BECHT & F. MILLER, supra note 8, at 211 (mentioning
the aggregate but-for test); Cole, supra note 7, at 810-11 (defendant liable if duplicative or
preemptive condition has same relation to plaintiff under the compensation rule as defendant does).
1985] CAUSATION IN TORT LAW 1781

are sufficiently "related" to the event to be treated as its causes. On its

own, the test cannot distinguish duplicative causes from preempted con-
ditions. Moreover, it would treat totally unrelated conditions as causes.
For example, consider the case in which C poisons P's tea and D
shoots P before P drinks the tea. Assume also that B was climbing Mt.
Everest. Clearly, D's shooting P was a preemptive cause of P's death,
and neither C's poisoning the tea nor B's climbing Mt. Everest was a
cause. The proposed test, however, would treat all three activities as
causes. In the aggregate they were a but-for cause of P's death, while
individually none of them was a but-for cause.

C. The Substantial-FactorFormula
Those theorists and judges who are unwilling to engage in contor-
tions such as the ones illustrated above to rescue the but-for test from its
inadequacies usually have turned to the substantial-factor formula, either
as a supplement to the but-for test (to apply to the overdetermined-causa-
tion cases) or as a complete substitute for it.
The substantial-factor formula was created originally not as a test of
actual causation but as a guide for resolving proximate-cause issues. The
originator of the formula, Jeremiah Smith, was content with the but-for
test as a test of cause-in-fact, with the usual exception for the overdeter-
mined-causation cases. 1 90 He wanted to devise a practical alternative to
the probability or foreseeability tests for determining the proximate-
cause limitations on liability, because he believed that those tests were
unsound and inconsistent with the results in many cases.19' He proposed
the following formula: "Defendant's tort must have been a substantial
factor in producing the damage complained of."192 ' The accompanying
explanation and alternative formulations clearly stated that the defend-
ant's tort could not be a substantial factor unless it satisfied the but-for
test (with an exception for simultaneous independent sufficient causes); in
addition, it would have to be an appreciable and continously effective or
efficient factor in producing the harm, up to the time of occurrence of the
193
harm.
Thus, the substantial-factor formula was meant to be used as the test
of legal (proximate) cause, but also incorporated the but-for test (and its
exception) for cause-in-fact. Smith's approach was adopted essentially

190. Smith, supra note 106, at 108-09 & n.20, 227, 312, 314 n.36, 316-17 n.41. Smith's
exception for overdetermined causation, however, was very narrow. It encompassed only a very
small subset of such cases-those in which there are two independent simultaneous tortious causes,
each by itself sufficient to produce the harm. Id. at 109 n.20, 312, 316-17 n.41.
191. Id. at 105, 115-28, 223-52, 308-09.
192. Id. at 309.
193. Id. at 310-12, 314 n.36.
1782 CALIFORNIA LAW REVIEW [Vol. 73:1735

intact in the original Restatement of Torts.194 It has persisted in the

Restatement (Second) of Torts' treatment of legal causation and cause-in-
fact, despite an attempt by Prosser and others to confine the substantial-
factor formula to the question of causation-in-fact.1 95
Leon Green first suggested that the substantial-factor formula be
applied to the actual-causation issue. He completely rejected the but-for
test, since it "take[s] the eye off the ball" by asking what would have
happened rather than focusing on what did happen and thereby denies
causation in many cases where causal contribution to the injury is obvi-
ous. 196 In his view, the substantial-factor formula was the best available
alternative. He argued that it should be used without elaboration to pose
the issue of causal contribution to the jury in every case, not just in the
overdetermined-causation cases. 197 Green's view was accepted by Pros-
ser, who passed it on to several generations of law students and lawyers
through the various editions of his hornbook, 198 even after Green himself
became disenchanted with the formula.199
The problem with the substantial-factor formula as a test of actual
causation (apart from its complete lack of guidance on what constitutes a
"factor") is that the alleged cause must be a substantial factor. Thus,
paradoxically, at the same time that Green was arguing strenuously and
cogently that the issue of causation should be completely divorced from
the policy issues camouflaged in proximate-cause analysis, he adopted a
formula for actual causation that required the judge or jury to determine
not only whether the actor's tortious conduct had contributed to the
injury (been a factor), but whether it had contributed enough to make the
actor responsible-whether, "in the light of all the other factors, the
defendant's conduct played an appreciable part in the result. ' '200 Green
seemed to believe that the resulting inquiry retained its character as a
factual, causal inquiry because it allegedly relied on quantitative rather
20
than qualitative measures of relative contribution. '
Even if quantitative measures of relative contribution are used, how-
ever, the question of limiting liability due to the extent of contribution,

194. RESTATEMENT OF TORTS §§ 431-435 (1934).

195. RESTATEMENT (SECOND) OF TORTS §§ 431-433 (1965). Compare W. PROSSER, supra note
1, § 42, at 248 (asserts formula confined to cause-in-fact issue) with A. BECHT & F. MILLER, supra
note 8, at 16, 130-34 (criticizes that assertion).
196. Green, supra note 8, at 556; see also L. GREEN, JUDGE AND JURY 192 (1930); L. GREEN,
supra note 2, at 167-70; Green, supra note 158, at 604-05; Green, supra note 8, at 557-59.
197. L. GREEN, JUDGE AND JURY 190-95 (1930); L. GREEN, supra note 2, at 137, 140, 181-85;
Green, supra note 158, at 603-07.
198. E.g., W. PROSSER, supra note 1, § 41, at 239-40; id. § 46, at 321, 323-24 (1st ed. 1941).
199. See infra text accompanying notes 206-08.
200. L. GREEN, supra note 2, at 134 (emphasis in original); see id. at 134-37, 141.
201. Id. at 140-41; see L. GREEN, JUDGE AND JURY 192-93 (1930); L. GREEN, supra note 2, at
84, 122.
1985] CAUSATION IN TORT LAW 1783

rather than due to the absence of any contribution, is clearly a proxi-

mate-cause issue of policy or principle, rather than an issue of actual
causation (contribution to the injury). Moreover, the substantial-factor
formula inherently invites consideration of qualitative measures as well
as quantitative ones.2 °2
Green himself, immediately after emphasizing the quantitative
nature of the inquiry, used qualitative considerations when discussing a
series of cases involving multiple causes. For example, he asserted that it
is "perhaps clear" that the hole poked by D's tug in P's barge, three feet
above the water line, was not an appreciable factor in the sinking of the
barge when compared with the continued loading of the barge which
brought the hole below the waterline, since those who were loading the
barge were fully conscious of what had occurred but made no effort to
repair the hole. Similarly, he suggested that there would be little possi-
bility of a finding that D's driving over a small flat box, which unbe-
known to him contained a highly explosive substance, was a substantial
factor in producing the resulting explosion. And, when discussing the
cases in which P is negligently or intentionally pushed by X into a hole
negligently maintained by D, Green stated that "what would be consid-
ered appreciable in one case might not in another. So, as compared with
X's negligence, [D 's] conduct might be considered appreciable, while as
compared with X's intended violence, [D's] conduct might not be so
considered."2 °3
In each of these cases, the defendant's tortious act was a necessary
condition for the occurrence of the injury. When an act (omission, con-
dition) was a necessary condition for the occurrence of an injury, it is
unnatural to deny that the act was, as a matter offact, not only a factor
in producing the injury, but also a substantial or appreciable factor. To
state otherwise is to make a noncausal, nonfactual policy judgment about
responsibility for the injury.
Prosser and, eventually, Green seem to have recognized the empti-
ness of the undefined substantial-factor formula and the danger of its
being used to introduce proximate-cause issues into the actual-causation
inquiry. Prosser, in the various editions of his hornbook, concluded his
brief discussion of the formula by equating it with the but-for test, with
an exception for the overdetermined-causation 20 5
cases, 2" and in his arti-
cles he relied primarily on the but-for test.

202. A. BECHT & F. MILLER, supra note 8, at 131-34; H.L.A. HART & T. HONORt, supra note
8, at 97, 293-94; C. MORRIS & C.R. MORRIS, supra note 4, at 174-75; Malone, supra note 6, at 88-97.
203. L. GREEN, supra note 2, at 153-58.
204. E.g., W. PROSSER, supra note 1, § 41, at 240 & n.33. The fifth edition of Prosser's
hornbook, edited by others, recognizes the policy judgments inherent in the substantial-factor
formula. PROSSER & KEETON, supra note 7, § 41, at 266-69.
205. E.g., Prosser, Proximate Cause in California,38 CALIF. L. REV. 369, 375-78 (1950).
1784 CALIFORNIA LAW REVIEW [Vol. 73:1735

In his later writings, Green abandoned his earlier substantive inter-

pretation of the substantial-factor formula. He treated it as a meaning-
less label for the judge to attach to her determination on the actual-
causation issue or to use as a hortatory device "to caution a jury to weigh
the evidence carefully."2 °6 He also criticized the view that the causal
inquiry involves not only the question of contribution to the injury, but
also the question of how significant the contribution was, without noting
that he previously had been the major proponent of this view.20 7 How-
ever, he continued to reject the but-for test and to insist that the issue of
causal contribution be submitted to the jury without any elaboration or
guidance.2 °s

D. Undefined, Directly Observable Causal Contribution

Lacking any satisfactory test, Green and a few other writers have
treated causation as an undefinable and irreducible factual relation
between events that can be directly perceived or inferred without explic-
20 9
itly or implicitly resorting to any generalization, definition, or test.
Most of these writers do not attempt to explain how this concept would
be applied in the various types of causation cases. However, there is one
notable exception: Becht and Miller's book, The Test of Factual
Causation.2 t°
Becht and Miller's book is a comprehensive attempt to demonstrate
how the concept of undefined, directly observable causal contribution is
applied in tort analysis. They argue that we determine whether a causal
relation exists between an act and an injury by breaking down the
sequence of events to the appropriate level of detail to see if we can "per-
ceive" a causal connection. If we can, the act is deemed to be a cause of
the injury even if the injury can also be traced back to another act or
event through a different causal sequence as in the merged-fires case.1
However, when Becht and Miller attempt to show how this approach
would work in the myriad variety of actual-causation cases, they are
forced to rely on the but-for test in an increasingly wide range of cases.
Initially, they are forced to use the but-for test to handle causal
inquiries involving omissions. By definition an omission is a nonevent-
something which did not happen-which only rarely will trigger an

206. Green, supra note 8, at 554.

207. Id. at 555; see id. at 557.
208. Id. at 548, 553-58.
209. A. BECHT & F. MILLER, supra note 8, at 9-10, 163-64; L. GREEN, supra note 2, at 132-33,
135, 137, 139-40, 183-85; Green, supra note 8, at 549, 553 & n.22, 554 n.25, 556, 560; see Borgo,
supra note 11, at 434-38; Epstein, supra note 13, at 161-63, 165-67; cf Williams, supra note 21, at 66-
68 (causal generalizations not involved in causal inquiry in the legal context).
210. A. BECHT & F. MILLER, supra note 8.
211. Id. at 9-11, 14-15, 18.
1985] CA USA TION IN TORT LAW 1785

actual causal sequence that can be directly perceived or traced. Instead,

an omission is almost always a "negative" cause of an injury: it repre-
sents a missed opportunity to prevent the injury. That is, an omission is a
cause of an injury only if the omitted act would have been part of a
hypothetical causal sequence that would have prevented, terminated, or
deflected the different causal sequence that actually occurred and pro-
duced the injury. Thus, in order to determine whether an omission was a
cause of an injury, it is necessary to conduct a hypothetical inquiry. The
omitted act must be hypothetically supplied, and a hypothetical causal
sequence (which Becht and Miller call a "parallel series") must be con-
structed and traced to determine whether it would have prevented the
occurrence of the injury.2 12
The hypothetical inquiry (but-for test) also must be used whenever
an actual ("positive") causal sequence becomes a negative cause by
blocking or terminating some preventive effort-for example, by block-
ing a fire engine or cutting off its supply of water.2 " 3 This inquiry also
must be employed to determine whether the excess (negligent) portion of
a certain speed or weight, or the difference between being in a safe posi-
tion or an unsafe (negligent) position, contributed to the injury.214
Indeed, as Becht and Miller reluctantly admit, the hypothetical inquiry
must be used whenever the negligent aspect of an act does not encompass
the act as a whole. For example, in the excess-speed case, the act is driv-
ing at a certain speed. We observe the car at that speed hit the child.
The negligent aspect of the speeding is the excess speed, which Becht and
Miller note "is not an event but a concept. Like an omission, it did not
happen, and it cannot be a simple [perceived] cause of events in the
world."2 '
Becht and Miller recognize the possible criticism that they are
implicitly using the hypothetical inquiry with all negligent acts, rather
than just with negligent acts that have a nonnegligent aspect, "but are
unconscious of [doing so] when the [hypothetical inquiry] shows clearly
that the negligent act was a cause of the harm."2' 16 They admit that they
cannot disprove this contention, but they continue to claim that they are
using direct perception of causation in such cases. However, they have

212. Id. at 21-24. Becht and Miller assert that an omission can only be a negative cause-that it
can only contribute by failing to prevent or block the positive causal sequence that actually produces
the result, rather than by itself triggering the actual causal sequence that produces the result. Id. at
170-71. But an omission can trigger an actual causal sequence. For example, an omission to salute
causes a soldier to be disciplined, or an omission to put money in the parking meter causes the
driver's car to be ticketed. The mental perception of the omission provides the connecting link.
213. Id. at 41-42.
214. Id. at 52-63.
215. Id. at 54; see also id. at 33, 52, 63, 65, 140.
216. Id. at 66.
1786 CALIFORNIA LAW REVIEW [Vol. 73:1735

some difficulty with a case in which the defendant negligently welds the
roof of a car, leaving a sharp edge, and the driver's head is crushed
against the roof at the weld point when the roof is smashed in during an
accident. They steadfastly assert that the case is one in which direct per-
ception or inference, unaided by any hypothetical inquiry, shows a lack
of causation by the sharp weld, just as direct perception or inference
shows that the color of the roof was not a cause of the injury.2 17
The direct-perception argument relies on the notion that there are
observable and describable causal qualities or forces in objects or events.
Becht and Miller eventually concede that this notion is naive as com-
pared with the Humean account, which is based on causal generaliza-
tions. 1 8 Even if the direct-perception notion were creditable, it is
difficult to imagine how the lack of causation by the sharp weld could be
perceived in the circumstances given. It is also difficult to understand
what is meant by "direct inference." An inference is based on some con-
cept of how things generally happen-that is, on causal generalizations.
Causal generalizations incorporate the belief that the cause is in some
sense necessary for the occurrence of the consequence. The but-for test is
simply the means by which we determine whether this element of neces-
sity exists in the particular case.
Becht and Miller argue that, even if it is true that they are explicitly
or implicitly using the but-for test in every case, they are using a less
restrictive version. Unlike the usual version, they compare what hap-
pened with what would have happened in as minute detail as possible.
Moreover, in those cases in which there was more than one positive
causal sequence that led to the injury, they only require that the defend-
ant's act or omission have been necessary for the continuation of the pos-
itive causal sequence with which it was associated, rather than requiring
that it have been necessary for the injury itself.2" 9
The first distinction-their use of minute detail-is nullified by their
tactic of "equating the injuries" to avoid finding but-for causation when-
ever there is a slight difference between the actual process or result and
the hypothesized process and result.22 °
The second distinction, however, is a real one. It enables them to
find that causation exists in many of the overdetermined-causation cases.
For example, in the merged-fires case the defendant's tortious conduct
will be treated as a cause of the destruction of the house if it was a but-for
cause of one of the fires, even though the house would have been burnt

217. Id. at 66-67; see also id. at 92, 139.

218. Id. at 163-64; see also infra text accompanying notes 228-31; T. BEAUCHAMP & A.
ROSENBERG, supra note 21, at 80-84, 107-17.
219. A. BECHT & F. MILLER, supra note 8, at 65-67; see also id. at 19, 31, 64-65.
220. Eg., id. at 28-31, 55, 58-59, 62, 64, 70-71, 91; see supra text accompanying notes 179-85.
1985] CAUSATION IN TORT LAW 1787

down by the other fire anyway. The defendant's tortious conduct was a
but-for cause of one of the duplicative causal sequences, and that is suffi-
cient to establish the causal connection to the destruction of the house.22 1
Becht and Miller's causal-contribution analysis ends up being very
similar to the "necessary element of a sufficient set" test. 222 However, it
is not exactly equivalent. Indeed, it produces anomalous results precisely
to the extent that it varies from the latter test. Becht and Miller's
approach requires that the tortious act or omission have been a necessary
condition for one of the positive (actual) causal sequences that contrib-
uted to the injury. They are therefore forced to deny that causation
exists when the overdetermined-causation problem results from duplica-
tive or preempted conditions that affect the same positive causal
sequence, rather than from the existence of multiple positive causal
sequences.
For example, they discuss the case in which C negligently fails to
discover and repair defective brakes in a car before renting the car to D,
who, unaware of the defective condition of the brakes, negligently fails to
try to use them to avoid hitting P. P is hit, but he would not have been
hit if the brakes had been working properly and D had used them. C's
omission (the failure to inspect and repair the brakes) was not a neces-
sary condition for the occurrence of the positive causal sequence that
produced the injury, since D made no attempt to use the brakes. Simi-
larly, given C's omission, D's omission (the failure to try to use the
brakes) was not a necessary condition either. Becht and Miller unhap-
pily but firmly conclude that neither omission was a cause of the injury,
but would hold both C and D liable to avoid a "morally indefensible"
and "inexcusable" result.223
The same problem occurs with acts. Assume that a cable has a max-
imum safe load capacity of one ton when it is in good condition. C negli-
gently weakens the cable (for example, by cutting a few of its strands) so

221. A. BECHT & F. MILLER, supra note 8, at 32.

222. Becht and Miller refer to and approve Hart and Honor6's "sufficient set" approach to
causation, but they avoid mentioning or discussing the "necessary element" aspect and in fact
erroneously imply that Hart and Honor6 do not believe that the notion of necessity is relevant to
causation. Id. at 202-04, 222. For a discussion of Hart and Honor6's position, see infra text
accompanying notes 228-32, 235-38, 252-60, 282-83.
223. A. BECHT & F. MILLER, supra note 8, at 95-98; see also id. at 82 n.116, 210-11. Prosser
reaches a similar conclusion:
It may be said with some confidence that if any such case is considered, both parties will be
held liable; but the theory of liability is not so clear. Perhaps the best guess is that each, by
his negligence, has deprived the plaintiff of a cause of action against the other, and so
should be liable.
W. PROSSER, supra note 1, § 41, at 239-40 n.25; accord J. FLEMING, supra note 4, at 174 n.19.
Green's treatment of the case is discussed supra text accompanying notes 108-15. For the resolution
of the case under the "necessary element of a sufficient set" test, see infra text accompanying notes
274-76.
 1788 CALIFORNIA LAW REVIEW [Vol. 73:1735

that it will now break if a one-ton load is applied. D negligently over-

loads the cable with a two-ton load, which would cause it to break even if
it were in good condition. The cable breaks at the weakened point, and
the load falls on and injures P. Since the negligent acts of C and D are
both part of the same positive causal sequence that produced the injury,
but neither was a necessary condition for the occurrence of the sequence,
neither is a cause of the injury under Becht and Miller's approach.22 4
Becht and Miller's book contains illuminating discussions on many
points-for example, the distinction between acts and omissions and the
need to focus on the tortious aspect of the actor's conduct. Perhaps its
most valuable contribution, however, is an unintended one: the demon-
stration of the unavoidability of the but-for (necessary condition) test in
causal analysis. This unavoidability is most clear when the causal
inquiry focuses on omissions or the tortious aspect of acts, but it is
implicit in all causal inquiries.2 25

E.The NESS (Necessary Element of a Sufficient Set) Test

In this section, I turn at last to a test for causal contribution that is
applicable to the entire spectrum of causation cases. This test incorpo-
rates the indispensable notion of necessity, but subordinates it to the
notion of sufficiency. I call this test the NESS (Necessary Element of a
Sufficient Set) test.
The NESS test was first suggested by H.L.A. Hart and Tony
Honor& However, their brief exposition of this test was overshadowed
and distorted by their primary emphasis on proximate-cause issues.2 26
The test has received very little notice and even less acceptance in the
subsequent legal literature.22 7 One of the principal purposes of this Arti-

224. A. BECHT & F. MILLER, supra note 8, at 57-58 & n.85. See infra text following notes 245
& 250 for the disposition of the case under the "necessary element of a sufficient set" test.
225. This point, at least with respect to omissions and the tortious aspect of acts, was not missed
by Green, who was a lifelong critic of the but-for test. In his earlier writings, he had focused the
causal inquiry on the negligent aspect of the actor's conduct. After reading Becht and Miller's book,
however, he insisted that the "orthodox analysis" limits the causal relation issue to the question of
whether the actor's conduct as a whole contributed to the injury. Green, supranote 8, at 546, 549-
51. He also claimed that there is no need to distinguish omissions from acts since "omissions to act
are merely incidents in a longer line of affirmative conduct and are easily resolved into it." Id. at
546; see id. at 546-47. We have already seen that this approach produces anomalous results unless
the tortious-aspect causation requirement is smuggled back in as part of the duty or proximate-cause
analysis. This is, in fact, what Green and his followers have done, with mixed success. See supra
text accompanying notes 111-26.
226. H.L.A. HART& A. HONORt,supra note 10, at 105-10, 116-19, 122, 216-29; H.L.A. HART
& T. HONORS, supra note 8, at 111-17, 122-25, 128-29, 235-53.
227. Fleming incorporated much of Hart and Honor6's exposition, including its deficient
aspects, into the second edition of his treatise. J. FLEMING, THE LAW OF TORTS 178-81 & n.h (2d
ed. 1961). The exposition has been retained with slight modifications in subsequent editions. E.g., J.
FLEMING, supra note 4, at 170-71, 173-74. Professors Cole, Becht, and Miller omit the necessary-
condition element and treat the test as a "sufficient set" approach. A. BECHT & F. MILLER, supra
1985] CAUSATION IN TORT LAW 1789

cle is to draw renewed attention to the NESS test, by demonstrating its

fundamental identity with the basic concept of causation and its utility in
resolving the cases that have proved troublesome for several generations
of tort scholars.

1. The PhilosophicBasis of the NESS Test

The NESS test captures the essential meaning of the concept of cau-
sation. This meaning was first articulated by the philosopher David
Hume. Hume rejected the earlier notion that we acquire causal knowl-
edge through direct sensory perception of causal qualities or forces inher-
ent in objects or events. Instead, he insisted that we only observe certain
successions of events, more or less frequently repeated. From these
observations, we inductively derive the belief that certain antecedent
events are not only always conjoined with, but also are sufficient for the
occurrence of, certain subsequent events. That is, we form a belief in
more or less well-grounded causal laws or generalizations. Causal
knowledge or belief also can be based on reasoning by analogy from such
observations or can be acquired through education, which passes on the
causal knowledge of others. Any singular causal statement about a par-
ticular occurrence is not a reporting of direct sensory perception of
causal forces, but rather an assertion of the belief that the occurrence
instantiates one or more causal laws or generalizations.2 28
A fully specified causal law or generalization would state an invaria-
ble connection between the cause and the consequence: given the actual
existence of the fully specified set of antecedent conditions, the conse-
quence must follow. In other words, the fully specified set of antecedent
conditions is sufficient for the occurrence of the consequence. In the typ-
ical singular causal statement, the causal assertion includes, explicitly or
implicitly, only a few of the antecedent conditions but nevertheless
asserts that they were part of an incompletely specified (and incompletely
understood or known) set of actual conditions that was sufficient for the
occurrence of the consequence.2 29

note 8, at 202-04, 222, discussedsupra note 222; Cole, supranote 7, at 485, 486 n.80, 488, 495, 505 &
n.133, 769, 770 n.16. Professors Fraser and Howarth also mention the NESS test, but attribute it to
Professor Papineau rather than to Hart and Honor6 and reject it, after misapplying it, in favor of a
probabilistic increased-risk "causal" test. Fraser & Howarth, More concern for cause, 4 LEGAL
STUD. 131, 133 & n.19, 135-38, 140-42, 145-56 (1984). Similarly, Mario Rizzo converts the INUS
(Insufficient but Necessary element of an Unnecessary but Sufficient set) test, which is similar to the
NESS test and was elaborated by Mackie subsequent to Hart and Honor6's book, into a
probabilistic-linkage test. Rizzo, supra note 178, at 1009-16; see infra text accompanying note 234.
228. T. BEAUCHAMP & A. ROSENBERG, supra note 21, at 4-11, 80-81, 87-88, 91-92, 139-43;
H.L.A. HART & T. HONORt, supra note 8, at 10-11, 14-15.
229. T. BEAUCHAMP & A. ROSENBERG, supra note 21, at 23-24, 84-87, 92-97, 114-15, 132, 136-
37, 141; H.L.A. HART & T. HONORt, supra note 8, at 10-11, 44-49; Mackie, supranote 21, at 23-24,
27-32, 35-38.
1790 CALIFORNIA LAW REVIEW [Vol. 73:1735

Hume was primarily interested in elaborating and defending this

"regularity" account of causation and causal knowledge, which insists
that the meaning of causation is to be found in causal laws or generaliza-
tions rather than in some notion of causal qualities or forces. However,
while pursuing this primary objective, he occasionally noted that one
must be careful when constructing a causal law or generalization to dis-
tinguish the causally relevant antecedent conditions from the causally
irrelevant antecedent conditions. This differentiation is necessary to
insure that the set of jointly sufficient antecedent conditions includes only
those that are indeed invariably connected with the consequence. Thus,
the antecedent conditions must be restricted to those that are necessary
for the sufficiency of the set.23 °
Hume also maintained that a certain consequence is always pro-
duced by the same cause-that is, that there is a unique sufficient set of
antecedent conditions that always must be present to produce a particu-
lar consequence. Hume's successor, John Stuart Mill, disagreed. He
asserted that there may be a plurality of potential causes for any conse-
quence. For example, death may be caused by poison on one occasion,
by a bullet wound on another, by fire on yet another, and so forth.
Hume's defenders have argued that such seemingly different causes may
upon further investigation turn out to be overly broad specifications of a
single common cause, or that the seemingly identical consequence may
turn out to be distinct consequences. However, they also admit that
ordinary experience provides strong support for the plurality-of-poten-
tial-causes theory. Indeed, the plurality theory has become part of the
dominant "regularity" account of the meaning of causation. 231
As Hart and Honor6 noted in 1959, the "necessary element of a
sufficient set" (NESS) test of causal contribution follows directly from
this dominant "regularity" account of the meaning of causation.2 32 The
essence of the concept of causation under this philosophic account is that
a particularcondition was a cause of (condition contributingto) a specific
consequence if and only if it was a necessary element of a set of antecedent
actualconditions that was sufficientfor the occurrence of the consequence.
(Note that the phrase "a set" permits a plurality of sufficient sets.) This
is the more precise, extended statement of the NESS test.
Surprisingly, however, this notion of a contributing condition
appears only infrequently in the philosophic literature on causation. The
literature continues to be characterized by arguments based on necessary

230. T. BEAUCHAMP & A. ROSENBERG, supra note 21, at 94-96, 114-17.

231. Id. at 23, 88-91; H.L.A. HART & T. HONORf, supra note 8, at 19-20; see id. at xxxix-xlii.
232. H.L.A. HART & T. HONOR(, supra note 8, at 111-18. Sosa incorrectly characterizes Hart
and Honor6's account as a simple sufficient- condition account. Sosa, supra note 170, at 1; see A.
BECHT & F. MILLER, supra note 8, at 202-04, 222, discussedsupra note 222.
1985] CAUSATION IN TORT LAW 1791

conditions, sufficient conditions, or necessary and sufficient conditions.23 3

After Hart and Honor6's book was published, several nonlegal philoso-
phers articulated NESS-like notions of causal contribution. In each
instance, however, they burdened the NESS test with the additional
requirement that the sufficient set be necessary for the occurrence of the
consequence in the particular circumstances. This requirement converts
the NESS test to a necessary and sufficient condition test, which cannot
accomodate the duplicative-causation cases (for example, the merged-
fires cases).2 34
Hart and Honor6 do not make this mistake, because they recognize
that there may be a plurality of causes in a particular instance. However,
their version of the NESS test is also deficient in several respects. First,
they apparently require that a duplicative or preemptive cause be inde-
pendently "sufficient" for the occurrence of the injury. That is, it must
be sufficient in conjunction with the "background" conditions, but
excluding the duplicated or preempted conditions. Second, they sub-
merge the critical distinction between duplicative and preemptive causa-
tion by constructing an overlapping typology of overdetermined-
causation cases. Finally, they confuse the factual issue of causal contri-
bution with the issues of policy or principle involved in determining ulti-
mate liability in certain types of cases.

2. The Duplicative-CausationCases
As elaborated, Hart and Honor6's NESS test is only slightly less
restrictive than the Restatement's substantial-factor exception to the but-
for test, which only applies when each of two actively operating factors
was sufficient by itself for the occurrence of the injury. 2 35 Hart and
Honore do not require that each of the factors have been actively operat-
ing, but they seem to require that each have been sufficient by itself for
the occurrence of the injury. 236 For example, in the merged-fires cases,
they assume that each fire would have been sufficient by itself for the
destruction of the plaintiff's house. The NESS test then confirms causal
contribution by each fire. Each fire was necessary for the sufficiency of a
set of actual antecedent conditions that did not include the other fire.23 7
The requirement that each factor have been sufficient by itself (when
combined with the background conditions) is too restrictive and is not a
part of the basic concept of causation that is reflected in the NESS test.

233. Sosa, supra note 170, at 1-3.

234. Mackie, supranote 21, at 16-21, 25-27 (also discussing Marc-Wogau's and Scriven's similar
accounts); see T. BEAUCHAMP & A. ROSENBERG, supra note 21, at 212-13; Sosa, supra note 170, at
3-4.
235. RESTATEMENT (SECOND) OF TORTS § 432(2) (1965).
236. H.L.A. HART & T. HONORt, supra note 8, at 123-24, 125, 206-07, 235-39, 245, 249.
237. Id. at 122-23, 206, 235, 237-39.
1792 CALIFORNIA LAW REVIEW [Vol. 73:1735

Moreover, the independent-sufficiency requirement is not followed by the

courts. For example, in two duplicative-causation cases involving
merged fires and noisy motorcycles, respectively, the courts did not
require the plaintiff to prove the independent sufficiency of each contrib-
uting factor, but rather required him to prove only that each factor con-
tributed to the injury.2 38 Similarly, in the pollution cases, the courts have
allowed the plaintiff to recover from each defendant who contributed to
the pollution that caused the injury, even though none of the defendants'
individual contributions was either necessary or sufficient by itself for the
occurrence of the injury.2 3 9
Malone argues that the courts' departure from the but-for test in the
overdetermined-causation cases is further evidence of the policy-depen-
dent nature of the causal inquiry. He contends that the courts are willing
to replace the but-for test with the substantial-factor formula in cases like
the merged-fires cases because of the important policies that underlie
"the well-established rules that prohibit the setting into motion of a
' ' 24
destructive force. 0
But, contrary to Malone's assertion, the policies or principles at
work in these cases are not those underlying the particular rule that was
violated. They are the ones underlying the actual causation requirement
itself. As Malone himself notes, the courts feel impelled to depart from
the but-for test in these overdetermined-causation cases because,
although the but-for test is not satisfied, it is clear that the defendant's
tortious conduct 241 contributed to the injury: "Our senses have told us
that he did participate. . . . In the language of the layman, the defend-
ant's fire 'had something to do with' the burning of the plaintiff's prop-
erty."24 2 Just as Newtonian mechanics serves as an adequate substitute
for the more accurate and comprehensive theories of relativity and quan-
tum mechanics in ordinary physical situations, the but-for test serves as
an adequate substitute for the NESS test in ordinary causal situations. In
each context, however, the substitute must give way to the more accurate
and comprehensive concept when the situation is more subtle and
complex.
In the pollution cases, the NESS test confirms that each defendant's

238. Corey v. Havener, 182 Mass. 250, 65 N.E. 69 (1902) (noisy motorcycles); Anderson v.
Minneapolis, St. P. & S. Ste. M. Ry., 146 Minn. 430, 179 N.W. 45 (1920) (merged fires). The cases
are discussed in Malone, supra note 6, at 90-91.
239. E.g., Michie v. Great Lakes Steel Div., 495 F.2d 213 (6th Cir.), cert. denied, 419 U.S. 997
(1974); Warren v. Parkhurst, 45 Misc. 466, 92 N.Y.S. 725 (1904), afid, 105 A.D. 239, 93 N.Y.S.
1009 (1905), aff'd, 186 N.Y. 45, 78 N.E. 579 (1906); see J. FLEMING, supra note 4, at 176.
240. Malone, supra note 6, at 89.
241. The term "tortious conduct" should always be interpreted to mean the tortious aspect of
the actor's conduct. See supra text accompanying notes 135-45.
242. Malone, supra note 6, at 89 (emphasis in original).
1985] CA USA TION IN TORT LAW

pollution contributed to the injury, even though it was neither necessary

nor independently sufficient for the injury. For example, assume that five
units of pollution were necessary and sufficient for the injury and that
each of seven defendants discharged one unit of pollution. Each defend-
ant can truthfully say that its one unit was neither necessary nor indepen-
dently sufficient for the injury. But each defendant's one unit was
necessary for the sufficiency of a set of actual antecedent conditions that
included only four of the other units, and the sufficiency of this particular
set of actual antecedent conditions was not affected by the existence of
two additional duplicative units.24 3
A similar causal situation exists even if one defendant discharges
five units of pollution and a second defendant discharges two units. The
two units still mix with the five units to produce the injurious seven units.
More rigorously, the two units were necessary for the sufficiency of a set
of actual antecedent conditions that included only three of the first
defendant's five units, a set whose sufficiency was not affected by the
existence of two additional duplicative units also provided by the first
defendant.
The analysis of the merged-fires cases is analogous. Thus, if any two
of three fires were sufficient for the injury, but none by itself was suffi-
cient, each was a cause of the injury since each was necessary for the
sufficiency of a set of actual antecedent conditions that included only one
of the other fires. The same causal situation exists even if there were only
two fires, one of which was independently sufficient and the other of
which was not. The first fire was clearly a cause, since it was indepen-
dently sufficient. But the second fire also was a cause. It was necessary
for the sufficiency of a set of actual antecedent conditions which included
another fire (the first) that was "at least large enough to be sufficient for
the injury if it merged with a fire the size of the second fire." The suffi-
ciency of this set is not affected by the fact that the first fire was so large
that it would have been sufficient by itself.
The wording of the quoted condition, "at least large enough," is not
a verbal gimmick. The condition is an actual one that existed on the
particular occasion. It describes a certain factual situation, as much as
the condition in the pollution case that referred to only three of the first
defendant's five units of pollution. In the former case, the size of the first

243. I assume that the injury was not accelerated or aggravated by the extra units of pollution.
If it was, causal contribution would be even clearer. I also assume that the units of pollution arrived
simultaneously at the site of the injury. Obviously, if five units arrived before the other two and
produced the injury before the other two arrived, the first five units were causes of the injury and the
last two were not. Their potential effects were preempted by the effects of the first five. See infra text
accompanying notes 247-51.
1794 CALIFORNIA LAW REVIEW [Vol. 73:1735

fire is broken down into portions; in the latter case, the amount of the
first defendant's pollution is broken down into portions.
The NESS test would attribute causal status to a very small fire that
merged with an overwhelming large fire. The very small fire was a dupli-
cative cause of any resulting injury. Whether the person who tortiously
caused the very small fire should be held liable for any or all of the result-
ing injury is an issue of policy or principle that comes under the heading
of damages. 2 "
The same analysis applies to the weakened-cable hypothetical.2 41 In
that example, a cable with a maximum safe load capacity of one ton was
weakened by C, who negligently cut a few of its strands so that it would
break if subjected to a one-ton load. D then negligently put a two-ton
load on the cable, which would have caused it to break even if the cable
were in good condition, and the cable broke at the weakened point. C's
weakening of the cable was necessary for the sufficiency of a set of actual
antecedent conditions which included a load of at least one ton, and the
sufficiency of this set was not affected by the fact that the load was more
than one ton. Indeed, the causal contribution of C's weakening of the
cable is evidenced by the fact that the cable broke at the weakened
point.2 46 Similarly, D's overloading of the cable was necessary for the
sufficiency of a set of actual antecedent conditions that included a cable
with a load capacity of at most one ton, and the sufficiency of this set was
not affected by the fact that the cable had a capacity of less than one ton
as a result of the weakening. C's weakening of the cable and D's over-
loading of it are duplicative causes of the cable's breaking and the result-
ing injury to the plaintiff.
3. The Preemptive-CausationCases
So far, the discussion has concentrated on the duplicative-causation
cases. I turn now to the preemptive-causation cases. Hart and Honors
mention situations in which the defendant asserts that someone else
would have shot the plaintiff or stolen his property if the defendant had
not done so.24 7 The defendant's argument would succeed under a literal
application of the Restatement, which adheres to the but-for test unless
the alternative factor was "actively operating."24 But, under the NESS
test, the defendant's tortious act clearly was a cause of the injury, since it
244. See infra text accompanying notes 258-73.
245. See supra text accompanying note 224.
246. Of course, if the cable broke at another point, Cs weakening of the cable would not be a
necessary element of any sufficient set of actual antecedent conditions. See infra text following note
250.
247. H.L.A. HART & T. HONORIf, supra note 8, at 124-25, 249-50.
248. RESTATEMENT (SECOND) OF TORTS § 432(2) (1965); see supra text accompanying note
235.
1985] CAUSATION IN TORT LAW

was a necessary element of the set of actual antecedent conditions that

was sufficient for the injury. Under this test, moreover, only the defend-
ant's actions would be deemed a cause. The potential actions of others
that did not in fact occur could not be a part of any set of actual antece-
dent conditions that was sufficient for the injury.
Preemptive causation also is present in many cases where the alter-
native factor actually did occur. For example, in my initial illustration
where D shot and killed P just as P was about to drink a cup of tea that
had been poisoned by C,24 9 D 's shot was necessary for the sufficiency of a
set of actual antecedent conditions that did not include the poisoned tea.
Conversely, C's poisoning of the tea was not a necessary element of any
sufficient set of actual antecedent conditions. A set that included the
poisoned tea but not the shooting would be sufficient only if P actually
drank the tea, but this was not an actual condition. The shooting pre-
empted the potential causal effect of the poisoned tea.
Moreover, even if P actually had drunk the poisoned tea, C's
poisoning of the tea still would not be a cause of P's death if the poison
did not work instantaneously but the shot did. The poisoned tea would
be a cause of P's death only if P drank the tea and was alive when the
poison took effect. That is, a set of actual antecedent conditions sufficient
to cause P's death must include poisoning of the tea, P's drinking of the
poisoned tea, and P's being alive when the poison takes effect. Although
the first two conditions actually existed, the third did not. D's shooting
P prevented it from occurring. Thus, there is no sufficient set of actual
antecedent conditions that includes C's poisoning of the tea as a neces-
sary element. Consequently, C's poisoning of the tea fails the NESS test.
It did not contribute to P's death.250
As the last example illustrates, a necessary condition for the suffi-
ciency of any set of actual antecedent conditions is that the injury not
have occurred already as a result of other actual conditions outside the
set. The determination of whether this condition existed, as with all the
other conditions, is an empirical judgment. For example, in the weak-
ened-cable hypothetical, C's weakening of the cable was considered to be
a duplicative cause when the cable broke at the weakened point. If the
cable had broken instead at some other point, our empirical evaluation
would be that the potential causal effect of C's weakening of the cable
was preempted by the existence of this other weaker point in the cable.
This other weaker point prevented the cable from staying together long
enough for C's weakened point to have any effect. In those circum-
stances, C's weakening of the cable would not be a necessary element of
any sufficient set of actual antecedent conditions.
249. See supra text following note 166.
250. See H.L.A. HART & T. HONORP, supra note 8, at 124.
1796 CALIFORNIA LAW REVIEW [Vol. 73:1735

The key to the overdetermined-causation cases, therefore, is the dis-

tinction between duplicative and preemptive causation. In each case, an
empirical judgment must be made: was the tortious aspect of the defend-
ant's conduct a necessary element in a set of actual antecedent conditions
that was sufficient for the occurrence of the injury, or was it not a neces-
sary member of any actually sufficient set because its potential effects
were preempted by other actual conditions outside the set?
In the merged-fires cases or the noisy-motorcycles case, we believe
that each factor duplicates or reinforces the effects of the other factor,
rather than preempting those effects-that is, that there is more than one
sufficient cause. In the shooting case, however, we believe that there was
only one sufficient cause (the shooting), which preempted the effects of
the other potential cause (the poisoning) and thereby prevented it from
being a necessary member of an actually sufficient set. We would believe
otherwise if medical experts testified that the effects of the poison com-
bined with the effects of the shooting to cause the victim's death,
although either alone would have been sufficient.2 51 The critical issue is
whether one factor completely preempted the potential effects of the
other factor, so that there was only one sufficient (preemptive) cause in
the actual circumstances, or whether each factor was a necessary part of
a different (but overlapping) actually sufficient set, in which case there
was more than one sufficient (duplicative) cause.
Hart and Honor6 submerge this critical issue in their discussion of
the overdetermined-causation cases. They construct a typology that first
divides the cases into those involving "additional causes" and those
involving "alternative causes." The alternative-cause category is meant
to include those cases in which there was a potential alternative cause
which never occurred. Examples include the cases where the defendant
asserts that someone else would have shot or robbed the plaintiff if he
had not.2 52 However, Hart and Honor6 also include a bridge-delay case
in which there was an actual second bridge that would have delayed the
plaintiff's boat even if the first bridge had not.2 3 It is misleading to
speak of these cases as alternative-cause cases. As Hart and Honor6
clearly indicate, the alternative cause was not a cause. It was a pre-
empted potential cause, best referred to as a "preempted condition." The
category is a subset of the preemptive-causation cases.
Their additional-cause category is subdivided into three subcatego-

251. Thus, the court's holding in State v. Scates, 50 N.C. (5 Jones) 420 (1858), that the
defendant could not be held liable, even though medical witnesses testified that his burning of a child
was the "primary cause" of the child's death, if an independent blow by someone else hastened the
death, is clearly erroneous unless the blow was an immediate, preemptive cause of the death. The
case is discussed in H.L.A. HART & T. HONOR, supra note 8, at 242.
252. H.L.A. HART & T. HONORt, supra note 8, at xl, 124-25, 207, 249-50.
253. I at 250-51.
1985] CA USATION IN TORT LAW 1797

ries. The "combinatory" or "reinforcing" cause subcategory, however, is

the only one that actually involves additional (duplicative) causes.2 54
The other two subcategories are subsets of the preemptive-causation
cases, where again it is misleading to refer to the preempted condition as
an additional cause. The "neutralizing" cause subcategory covers situa-
tions in which the effective cause prevents an alternative active causal
process from being completed. For example, a fire approaching a house
is quenched by a flood which destroys the house, or a poison victim is
killed by a bullet before the poison kills him.255 The "overtaking" cause
subcategory originally was limited to situations in which a subsequent
actual injury would have caused the same damage as the initial injury:
for example, two successive injuries, each of which would have been suffi-
cient to disable the victim totally.2 56 However, the subcategory has been
expanded in the second edition to encompass situations in which the
effective cause frustrates an active causal process, so that the subcategory
now overlaps the "neutralizing" cause subcategory.25 7
Perhaps as a result of their confusing typology, Hart and Honor6
lose sight of the basic concept of causation embodied in the NESS test.
In the double-bridge and successive-injury cases, for example, the
defendant's tortious conduct was a (preemptive) cause of the boat's delay
in the double-bridge case and a (preemptive) cause of the victim's loss of
a limb and resulting total disability in the successive-injury case. But
Hart and Honor6 claim that the tortious conduct was not a cause of the
victim's economic loss in either case, since the victim would have suf-
fered the economic loss anyway as a result of the second bridge or the
second injury, respectively.2 58
This causal argument, however, is based on the but-for test rather
than the NESS test. In each case, the defendant's tortious conduct was a
necessary element of a set of actual antecedent conditions that did not
include the subsequent condition (the second bridge or the second injury)
and was sufficient for the occurrence of the economic loss. The existence
of the subsequent condition did not prevent this set from being sufficient
to produce the economic loss. On the other hand, neither the second
bridge nor the second injury was a necessary element of any set of actual
antecedent conditions that was sufficient for the occurrence of the loss.
The second bridge will cause a delay only if the boat reaches it, but the
boat did not reach it.259 The second injury will cause total disability and
resulting economic loss only if the victim is not already totally disabled,

254. Id. at 237-39.

255. Id. at 124, 206-07, 239-45.
256. H.L.A. HART & A. HONORP, supra note 10, at 118, 190, 224-25.
257. H.L.A. HART & T. HONORt, supra note 8, at 245.
258. Id. at 246-48, 250-51.
259. See L. GREEN, supra note 2, at 168-70.
1798 CALIFORNIA LAW REVIEW [Vol. 73:1735

but this also was not one of the actual conditions. In each instance, the
defendant's tortious act is a preemptive cause of the injury and the result-
ing economic loss, and the subsequent preempted condition is not a cause
260
of the economic lOSS.

4. Distinguishingthe Damages Issue: The Successive-Injury and

Overwhelming-Force Cases
The successive-injury cases have engendered much debate and con-
fusion, particularly in the Commonwealth countries, where the legal
community seems unable to free itself from the but-for concept of causa-
tion. 261 The causal situation is clear in these cases. The first injury
caused the economic loss; the second did not. The issue is not causal. It
is a proximate-cause issue of policy or principle that is most appropri-
ately placed under the heading of damages, and it also arises in the dupli-
cative-causation cases. The issue is whether a defendant who has
tortiously caused injury to the plaintiff nevertheless should be absolved
from liability if the injury would have occurred anyway as a result of
independent duplicative or preempted conditions.
Courts generally absolve the defendant from liability if he proves
that the injury would have occurred anyway as a result of independent
nontortious conditions. 262 In such a case, the plaintiff's corrective-justice
claim-that he would not have been injured if not for the tortious con-
duct of others-fails. On the other hand, if the duplicative or preempted
conditions also resulted from tortious conduct, the plaintiff's corrective-
justice claim is intact.26 3 Nevertheless, it may seem unfair to hold a
defendant liable for the entire injury if his contribution to the injury was
relatively minor compared to the other contributing conditions. But this
proximate-cause problem has been greatly lessened by the recent wide-
spread liberalization of the rules governing contribution among defend-
ants.26 Therefore, once the plaintiff has established that the tortious
aspect of a certain defendant's conduct contributed to the injury, many
260. In the successive-injury case, the second condition is a cause of the second injury, and of
any additional pain or costs resulting from the second injury.
261. E.g., Jobling v. Associated Dairies, 1982 A.C. 794 (1981); J. FLEMING, supra note 4, at
174-76; Fraser & Howarth, supra note 227, at 136-37, 145-56; McGregor, Successive Causes of
PersonalInjury,33 MOD. L. REV. 378 (1970); Strachan, supra note 7; Williams, supra note 21, at 75-
79. For similar confusion by American writers, see Henderson, supra note 105, at 201-12; Peaslee,
supra note 175.
262. E.g., Kingston v. Chicago & N.W. Ry., 191 Wis. 610, 211 N.W. 913 (1927) (dicta); Cook v.
Minneapolis, St. P. & S. Ste. M. Ry., 98 Wis. 624, 74 N.W. 561 (1898); Jobling v. Associated Dairies,
1982 A.C. 794 (1981). But see RESTATEMENT (SECOND) OF TORTS § 432(2) comment d (1965);
Carpenter, Concurrent Causation, 83 U. PA. L. REV. 941 (1935).
263. Eg., Kingston v. Chicago & N.W. Ry., 191 Wis. 610, 211 N.W. 913 (1927); Baker v.
Willoughby, 1970 A.C. 467 (1969).
264. E.g., American Motorcycle Ass'n v. Superior Court, 20 Cal. 3d 578, 578 P.2d 899, 146 Cal.
Rptr. 182 (1978); Dole v. Dow Chem. Co., 30 N.Y.2d 143, 282 N.E.2d 288, 331 N.Y.S.2d 382
1985] CAUSATION IN TORT LAW 1799

courts shift the burden to the defendant to establish that (1) the injury
would have occurred anyway as a result of independent nontortious con-
ditions, (2) he contributed to only a part of the injury, or (3) he is entitled
to contribution from the other defendants based on relative tortious
contribution.2 65
The causal issue is almost always confused with the policy-laden
damages issue in the overwhelming-force cases involving passive condi-
tions. When an overwhelming force combines with an active condition,
the courts are less likely to confuse the two issues. Thus, in the merged-
fires cases, the courts generally recognize that the defendant's smaller fire
that combined with a much larger fire nevertheless contributed to the
injury. If the defendant avoids liability, it is not due to a lack of causal
contribution but rather due to noncausal limitations. For example, the
defendant usually can avoid liability if he proves that the other fire was
independently sufficient and of nontortious origin, or perhaps even if it
was also of tortious origin if it was so overwhelming as to make the
defendant's contribution relatively insignificant.26 6 On the other hand,
the defendant is unlikely to escape liability even though his contribution
was a very small part of the total cause if it was necessary for the conse-
quence or, although unnecessary, was combined with a large number of
similarly small contributions, as in the pollution cases.26 7
When the overwhelming force combines with a passive condition,
however, most courts and commentators tend to fall back on the but-for
test and erroneously deny causal contribution. 268 For example, the
courts deny that the defendant's negligent failure to construct an ade-
quate dam or to keep runoff pipes free of debris was a cause of flooding
that would have occurred given a normal storm, if the actual storm was
an extraordinary one that would have caused flooding even if the defend-
(1972); Landes & Posner, Joint and Multiple Tortfeasors: An Economic Analysis, 9 J. LEGAL STUD.
517, 550-51 (1980); Robinson, supra note 7, at 716 & n.12.
265. In addition to the cases cited supra notes 262-64, see Michie v. Great Lakes Steel Div., 495
F.2d 213 (6th Cir.), cert. denied, 419 U.S. 997 (1974); Maddux v. Donaldson, 362 Mich. 425, 108
N.W.2d 33 (1961). Some courts have adopted a "comparative causation" allocation formula in the
strict product liability cases, since there is no negligence or fault as is required under the
comparative-negligence (fault) rubric. E.g., General Motors Corp. v. Hopkins, 548 S.W.2d 344
(Tex. 1977). However, from a purely causal standpoint, each necessary (or NESS) condition is
equally a cause. At best, the comparative-causation formula is a camouflage for a "comparative
responsibility" approach (surely a better and more helpful term), in which the types and degrees of
tortious conduct involved and the policies underlying each type of tortious conduct are weighed to
allocate liability. E.g., H.L.A. HART & T. HONORt, supra note 8, at 232-34; see Robinson, supra
note 7, at 758-68. At worst, the comparative-causation formula invites resort to mechanistic,
generally inapplicable calculations of relative energy or force, which have little to do with the basic
principles and goals of tort liability. E.g., Epstein, Defenses, supra note 49, at 179-80; Epstein, Reply,
supra note 49, at 478 n.5.
266. E.g., Kingston v. Chicago & N.W. Ry., 191 Wis. 610, 615-16, 211 N.W. 913, 915 (1927).
267. See cases cited supra note 239.
268. This point is noted in Malone, supra note 6, at 92-94.
 1800 CALIFORNIA LAW REVIEW [Vol. 73:1735

ant had not been negligent.26 9 Similarly, they have stated that the
defendant's negligent failure to remove a rotten pole is not a cause of
injury when the pole is knocked down by a car that hits it with such force
that it would have been knocked down even if it were not rotten.270
These are overdetermined-causation cases that are entirely analo-
gous to the merged-fires cases and the weakened-cable case.27 t Thus, if
the rotten pole broke at its weak point, the weak condition clearly con-
tributed to the injury. It was a necessary element of a sufficient set of
actual antecedent conditions that included an impact with at least
enough force to knock down a rotten pole. The sufficiency of the set was
not affected by the fact that the impact had a greater force, unless the
impact knocked the pole over in its entirety rather than breaking it at its
weak point. Similarly, the failure to remove debris from a dam's runoff
pipes or to build it high enough to prevent flooding during a normal
storm is a duplicative cause of flooding that occurs during an extraordi-
nary storm. The unremoved debris or inadequate height is a necessary
element in a sufficient set of actual antecedent conditions that includes an
at least normal storm, and the sufficiency of this set is not affected by the
fact that the storm was larger than normal.
A few courts have recognized that the defendant's tortious conduct
contributed to the injury in these passive-condition cases. Indeed, some
courts have held the defendant liable even when the injury would have
occurred anyway as a result of independently sufficient nontortious con-
ditions 272 However, if the noncausal policy limitations adopted in the
merged-fires and successive-injury cases are followed, the defendant
should be able to avoid liability in such circumstances. These policy lim-
itations, rather than the false denial of causation, explain the ultimate
result in those cases in which liability is denied because the injury would
have occurred anyway due to the extraordinary force of a storm.
Conversely, given the usual policy limitations, the defendant should
not escape liability when the duplicative or preempted condition was also
of tortious origin-for example, when the defendant's rotten pole is
knocked down by a negligent driver and an innocent third party is
injured. This case is the same as a merged-fires case in which the defend-

269. E.g., City of Piqua v. Morris, 98 Ohio St. 42, 120 N.E. 300 (1918); RESTATEMENT
(SECOND) OF TORTS § 432(l) illustration 2 (1965); A. BECHT & F. MILLER, supra note 8, at 90-93;
L. GREEN, supra note 2, at 150-51; PROSSER & KEETON, supra note 7, § 41, at 265-66 & n.16.
Contra RESTATEMENT (SECOND) OF TORTS § 450; Malone, supra note 6, at 92-94.
270. E.g., Gibson v. Garcia, 96 Cal. App. 2d 681, 687, 216 P.2d 119, 123 (1950).
271. See supra text following notes 245 & 250.
272. O'Connor v. Chicago, M. & St. P. Ry., 163 Wis. 653, 654, 158 N.W. 343, 344 (1916) (weak
tree blown down by storm that would have blown down sound tree); Nitro-Phosphate & Odam's
Chem. Manure v. London & St. Katherine Docks, 9 Ch. D. 503, 527 (C.A. 1878) (levee built too low
overflowed by extraordinarily high tide that would have overflowed levee built to proper height); see
Carpenter, supra note 127, at 418-19.
1985] CAUSATION IN TORT LAW

ant's insufficient fire joins with an independently sufficient fire, also of

tortious origin. In each case, both conditions are tortious duplicative
causes of the injury, and liability should be imposed jointly on both
tortfeasors with a right of contribution.2 7 3

5. Theft, Nonuse, or Misuse of Defective or Missing Safety Devices

Some of the most difficult overdetermined-causation cases are the
preemptive-causation cases involving theft, nonuse, or misuse of defec-
tive or missing safety devices. A familiar example is the case in which C
negligently failed to discover and repair defective brakes in a car that he
rented to D, and D negligently failed to try to use the brakes to avoid
running into P. It is assumed that the injury to P would have been
avoided if and only if C had repaired the brakes and D had tried to use
them. As we saw above, Green asserts that C's conduct as a whole and
D's conduct as a whole were each causes of the injury, but that only C
should be held liable.274 Becht, Miller, Prosser, and Fleming assert that
neither C's negligence nor D's negligence was a cause of the injury, but
that both C and D should be held liable.2 7
Under the NESS test, it is clear that D's negligence was a preemp-
tive cause of P's injury, and that C's negligence did not contribute to the
injury. D's failure to try to use the brakes was necessary for the suffi-
ciency of a set of actual antecedent conditions that did not include C's
failure to repair the brakes, and the sufficiency of this set was not affected
by C's failure to repair the brakes. A failure to try to use brakes will
have a negative causal effect whether or not the brakes are defective. On
the other hand, C's failure to repair the brakes was not a necessary ele-
ment of any set of antecedent actualconditions that was sufficient for the
occurrence of the injury. Defective brakes will have an actual causal
effect only if someone tries to use them, but that was not an actual condi-
tion here. The potential negative causal effect of C's failure to repair the
brakes was preempted by D's failure to try to use them.2 76
The same analysis applies in every case involving nonuse or misuse
of a missing or defective safety device, unless the actor did not try to use
the device because he knew it was missing or defective.277 . For example,
failure to provide a fire escape was not a cause of P's injury if P could not
have gotten to the fire escape anyway or would not have tried to get to it

273. See supra text accompanying notes 262-67.

274. See supra text accompanying notes 108-15.
275. See supra note 223 and accompanying text.
276. E.g., Saunders Sys. Birmingham Co. v. Adams, 217 Ala. 621, 117 So. 72 (1928).
277. In the latter situation, the causal effect of the missing or defective device is a mental rather
than a physical one, which actually exists and is not preempted but rather completed by the actor's
failure to try to use the device. See supra notes 117 & 212.
1802 CALIFORNIA LAW REVIEW [Vol. 73:1735

even if it existed.278 Similarly, failure to provide safety equipment to2 79a

worker is not a cause of his injury if he would not have used it anyway,
and failure to pack a parachute properly is not a cause of death if the
deceased never pulled the ripcord to release the parachute.
A somewhat different situation is presented by McLaughlin's hypo-
thetical in which D empties P's water keg and refills it with salt before P
is dropped into the middle of a desert. C subsequently steals the keg,
thinking that it contains water, and P dies of thirst.2 80 This case is differ-
ent from the previous cases because it is assumed that P would have used
the water ("safety device") if it had been available. D's emptying the keg
made the water unavailable and was a preemptive cause of P's death.
C's theft of the keg was not a cause of P's death. Its potential effect
(making the water unavailable) depended on the keg's having water in it
when it was stolen, but that was not an actual condition.2 8 '
Hart and Honors modify the facts by having D poison the water in
the keg rather than replacing it with salt. They conclude that neither C
nor D was a cause of P's death, but only because they equate causing
death with shortening life and note that P would have died sooner by
poisoning if C had not stolen the keg.282 As Becht and Miller point out,
and as a NESS analysis demonstrates, the lack of water (fluid) caused the
death, and it is irrelevant that P would have died sooner if he had drunk
the poisoned water.28 3 He in fact did not drink the poisoned water,
because C stole it. C's theft caused P's death and preempted the poten-
tial causal effect of D's poisoning the water.
It should be clear by now that the NESS test not only resolves but
also clarifies and illuminates the causal issues in the problematic causa-
tion cases that have plagued tort scholars for generations. It does so
because it is not just a test for causation, but is itself the meaning of
causation. When there is no overdetermined-causation problem-that is,
when there is only one actual or hypothetical sufficient set of conditions
for a particular event-the NESS test collapses into the simple, tradi-

278. E.g., Weeks v. McNulty, 101 Tenn. 495, 48 S.W. 809 (1898). Contra Fraser & Howarth,
supra note 227, at 135-36.
279. E.g., M'Williams v. Sir Williams Arrol Ltd., 1962 Sess. Cas. 70 (H.L.); Qualcast
(Wolverhampton) Ltd. v. Haynes, 1959 A.C. 743; H.L.A. HART & T. HONOR9, supranote 8, at 127.
Contra Fraser & Howarth, supra note 227, at 135-36.
280. McLaughlin, supra note 175, at 155 n.25.
281. See supra text accompanying notes 250, 258-60. Becht and Miller conclude that neither C
nor D was a cause. A. BECHT & F. MILLER, supra note 8, at 210. This erroneous conclusion results
from their application of the but-for test to two factors that were both in the same causal sequence.
See supra text accompanying notes 222-24.
282. H.L.A. HART & T. HONORt, supra note 8, at 239-40 & n.70.
283. A. BECHT & F. MILLER, supra note 8, at 205-10. Becht and Miller rely on the tautological
"would not have died when and as he did" argument. Id. at 210; see supra text accompanying notes
176-78. The applicable NESS analysis is discussed supra text accompanying notes 249-50.
1985] CA USA TION IN TORT LAW 1803

tional but-for test. But the concept of causation is much more subtle and
complex than the but-for test. Thus, whenever the but-for test is not
satisfied, the factfinder must utilize this more complex notion of causa-
tion, with its often subtle distinction between actual sufficiency and mere
apparent (preempted) sufficiency, before the causal inquiry can be
concluded.

F The FactualNature of the CausalInquiry

Now that the content of the causal inquiry has been identified, we
are in a better position to evaluate the frequent claims that the causal
inquiry is dependent on policy considerations. In the first portion of this
section, I argue, contrary to the commonly held view, that the hypotheti-
cal inquiry associated with the necessary-condition criterion does not
introduce policy issues into the causal inquiry. In the second portion, I
examine and reject the various arguments made by Wex Malone, who
generally is thought to have proven that the causal inquiry is pervaded by
policy considerations. I conclude that the causal inquiry is a factual,
empirical inquiry that can be-and almost always is-kept distinct from
the policy issues in tort adjudication.

1. The Hypothetical-InquiryArgument
Many legal writers assert that the counterfactual analysis of neces-
sity "takes the eye off the ball" by asking what would have happened,
rather than focusing on what actually happened, and thereby injects pol-
icy issues into the causal inquiry. Some writers deplore this alleged effect
28 5
of necessity analysis,"8 4 while others embrace it.
The message of the preceding sections of this Part, however, is that
the notion of necessity is fundamental to the concept of causation,
although it is subsidiary to the notion of sufficiency rather than being the
direct and exclusive criterion of causal contribution as in the but-for test.
Efforts to escape from necessary-condition analysis by applying the
causal inquiry to the actor's conduct as a whole or resorting to une-
laborated formulas or paradigm cases are therefore ultimately futile.28 6
The causal question is not simply "What happened? ' 28 7 but rather
"How did it happen?" Since Hume, it has generally been accepted that
there is no inherent causal force or quality in objects which can be
directly observed. We observe only certain successions of events and

284. E.g., Green, supranote 158, at 605; Green, supra note 8, at 556-57, 559; Thode, supranote
100, at 426-27 & n.15, 431; Note, supra note 27, at 1531-33; see Epstein, supra note 13, at 160-61.
285. E.g., Malone, supra note 6, at 67-68 & n.9; sources cited supra note 7.
286. See supra text accompanying notes 49-99, 107-17 & 190-225.
287. E.g., W. PROSSER, supra note 1, § 41, at 237; Epstein, supra note 13, at 160; Morris, supra
note 3, at 1088-89; Thode, supra note 100, at 431, 433-34.
1804 CALIFORNIA LAW REVIEW [Vol. 73:1735

conditions, and we infer a causal relation in a particular succession if we

believe it instantiates an accepted causal law or generalization. Causal
generalizations, in turn, incorporate the notion of necessary condi-
tions.2 88 Thus, as legal and nonlegal philosophers have noted, the neces-
sary-condition criterion and its implicit counterfactual assertions are part
of the very meaning of causation.28 9
In tort adjudication, the causal inquiry is focused on the tortious
aspect of the actor's conduct. The trier of fact must determine whether
all the conditions that were necessary to make the defendant's conduct
tortious, including the relevant circumstances as well as the particular
act or omission, contributed to the injury. 290 The counterfactual analysis
of necessity serves two purposes in this causal inquiry. First, it reduces
the chance that a tortious condition erroneously will be treated as part of
a causal generalization that is believed to be applicable to the particular
injury. Second, it reduces the chance that preemptive causes will be
overlooked. Thus, the counterfactual analysis of necessity does not take
the eye off the ball. It rather helps to ensure that the right ball is being
used. The objection to counterfactual analysis of necessity per se is there-
fore misplaced.
Most objections to the counterfactual analysis of necessity have
focused on the clearly erroneous denials of causation that would be com-
pelled by the but-for test if it were applied to the overdetermined-causa-
tion cases. 2 91 This problem arises, however, only because the but-for test
makes the necessary-condition criterion the exclusive criterion for causal
contribution. The problem is eliminated under the NESS test, which
subordinates the necessary-condition criterion to the sufficiency criterion.
However, some writers have insisted that the counterfactual analysis
of necessity inherently depends on value judgments. Ironically, most of
these writers believe that causal analysis is a factual analysis. They there-
fore reject the counterfactual analysis of necessity because they think that
it introduces policy considerations into the causal analysis.2 92 They do
not realize that their arguments, if successful, would negate their claims

288. See supra text accompanying notes 228-30.

289. T. BEAUCHAMP & A. ROSENBERG, supra note 21, at 119-20; H.L.A. HART & T. HONORI,
supra note 8, at 14-15, 109-12; Mackie, supra note 21, at 30-31.
290. See supra text accompanying notes 135-45.
291. E.g., EPSTEIN, TORTS, supra note 4, at 271-72; Calabresi, supra note 9, at 86-87, 105-08;
Green, supra note 158, at 604-05; Green, supra note 8, at 556-60; Landes & Posner, supra note 9, at
109-11, 134; Malone, supra note 6, at 67, 88-97; Rosenberg, supra note 7, at 863-64; Thode, supra
note 100, at 431-33; Williams, supra note 21, at 75-79.
292. Cole, supra note 7, at 482-87, 505-06 & n.133, 768-74 & n.16; Epstein, supra note 13, at
160-61, 165-66; Thode, supra note 100, at 424-27, 431. CompareMalone, supra note 6, at 67-69 &
n.9 (welcoming policy considerations into causal inquiry), with id. at 71 & n.25 (causal issue usually
clear on the facts) and Malone, supra note 27, at 370-73 (causal issue always factual).
1985] CA USA TION IN TORT LAW 1805

that the causal inquiry is factual, since the counterfactual analysis of

necessity is an indispensable element of causal analysis.
The argument against counterfactual analysis is developed most
fully by Robert Cole. Initially, he asserts that the very decision to
employ one form of necessary-condition analysis rather than another is a
policy decision. He claims that there is no independent, policy-neutral
basis for choosing "necessary element of a sufficient set" (the NESS test),
"necessary on the particular occasion" (the but-for test), "necessary on
every occasion," or any other necessity formula.2 93 But, as we have
seen,29 4 the "necessary element of a sufficient set" formula is the essence
of the concept of causation. The other formulas do not capture the
notion of causal contribution, although one of them (the but-for test) is
an adequate proxy in most cases.
Cole also argues that the courts introduce policy considerations
when they assume, as part of the counterfactual situation, that the
defendant would have behaved reasonably. In particular, he argues that
when the courts analyze what would have happened if the defendant had
not been driving at eighty miles per hour, there is no neutral basis for
them to pick the legal speed limit (or some lower reasonable speed if the
legal speed was unsafe in the particular circumstances) as the basis for
the counterfactual inquiry. They should instead consider each possible
alternative speed that the defendant might have chosen other than eighty
miles per hour and the changes in behavior (such as greater attentiveness
or deference to other drivers) that might accompany each alternative
speed. In order to decide whether the defendant's eighty miles per hour
speed was a necessary condition for the accident, the court must consider
all these alternatives in as much detail as possible. It must then deter-
mine whether some speed other than eighty miles per hour was not only
noticeably more likely than all the other possible alternative speeds but
also more likely to result in the accident. If such an alternative speed
exists, the eighty miles per hour speed was not a necessary condition for
the accident.2 95
Cole then notes that the "pyramiding of conjectural possibilities"
inherent in any such detailed consideration of what might have happened
if the defendant had not been driving at eighty miles per hour makes it
impossible to reach any definite conclusion.2 96 When the courts select
the legal speed as the counterfactual speed and do not consider behav-
ioral changes that might be expected from someone driving at the legal

293. Cole, supra note 7, at 769-71 & nn.15 & 16.

294. See supra text accompanying notes 228-32.
295. Cole, supra note 7, at 769-70, 773-76, 783-84; see also Thode, supra note 100, at 426-27 &
n.17.
296. Cole, supra note 7, at 776-77.
1806 CALIFORNIA LAW REVIEW [Vol. 73:1735

speed rather than at eighty miles per hour, they are using policy consid-
erations to drastically reduce the range of conjecture about what would
have happened if the defendant had not been driving at eighty miles per
297
hour.
This argument depends on the common but fallacious assumption
that the courts are trying to determine what would or might have hap-
pened if the defendant had not been driving at eighty miles per hour.
Insofar as the causal inquiry is concerned, the courts are not interested,
and should not be interested, in that broad inquiry. They are interested
only in determining the causal effect of the tortious aspect of the defen-
dant's conduct. For example, in the speeding cases the courts want to
know whether the excess speed (the tortious aspect) was a necessary ele-
ment in some set of actual antecedent conditions that was sufficient for
the occurrence of the accident. They therefore change only the tortious
condition when setting up the counterfactual situation. They remove the
speed in excess of the legal speed from the set of actual antecedent condi-
tions to determine whether the set would have been sufficient without it.
They want to know whether the tortious aspect (the excess speed) actu-
ally contributed to the accident. They do not want to know, as part of
the causal inquiry, whether the defendant would have driven at the legal
speed if he had not driven at eighty miles per hour, or whether he might
have been more attentive or otherwise more careful if he had been driv-
ing at the legal speed. The causal issue is restricted to the narrow ques-
tion of the actual effect of the excess speed given the other conditions
(attentiveness, etc.) that actually existed.
Similarly, the courts do not change the actual conduct of the plain-
tiff or third parties when they set up the counterfactual situation. They
are not using policy considerations, but rather careful causal methodol-
ogy, when they remove only the defendant's excess speed in order to
determine the causal effect of that excess speed in the actual
298
circumstances.
Of course, once a court sets up the counterfactual situation by elimi-
nating the particular tortious condition, it must then determine the hypo-
thetical consequences of that change. As even Malone admits, 99 in most
cases there is little difficulty, for example, when the change is removing
the explosive character of a substance or the act of firing a gun. The
analysis becomes more complicated when human reactions to the
changed situation must be estimated. Again, however, the analysis is
usually fairly easy and not too speculative. People's reactions generally
will be fairly predictable using causal generalizations in which there is a

297. Id. at 771, 784-85, 792-97; see also Malone, supra note 6, at 67.
298. Contra Thode, supra note 100, at 426 n.15.
299. Malone, supra note 6, at 71 & n.25.
1985] CAUSATION IN TORT LAW 1807

high degree of confidence. Indeed, the defendant normally will stipulate

that he would have reacted in the usual manner-for example, by apply-
ing his brakes if he would have had time to do so at the legal speed after
seeing the plaintiff in his path. As Cole himself notes, it does not help the
defendant's case to argue that one specific act of negligence did not con-
tribute to the injury because he would have caused it anyway by some
other negligent act or omission.30 0
In any event, the hypothetical inquiry is not policy-driven. Instead,
empirically derived causal generalizations are used to estimate the most
likely consequences of the counterfactual situation. Cole seems to believe
that causal generalizations are not used in the counterfactual inquiry, or
that the hypothetical development of the counterfactual situation must
be specified in such detail as to make its occurrence (in that precise man-
ner) improbable. 30 ' As philosophers have noted, however, counterfac-
tual assertions are nothing more than "telescoped arguments" based on
more or less well-grounded causal generalizations. The credibility of a
particular counterfactual assertion depends on the extent of empirical
support for the causal generalizations themselves and the extent to which
the conditions that make up the counterfactual setting correspond to
those that are specified in the causal generalizations. 302 These condi-
tions, as Hart and Honor6 have argued, must be specified at a fairly high
level of generality if they are to serve as a broadly applicable causal gen-
eralization, rather than as a detailed unique description of a particular
event with no causal explanatory force.3 ° 3

2. Malone's Arguments
Malone attacks the factual nature of the causal inquiry differently.
He does not challenge the process by which the counterfactual setting is
established, and he recognizes that causal generalizations provide the
foundation for any particular causal judgment. Instead, he presents an
array of arguments to demonstrate that the policies underlying particular
legal rules affect both causal judgments themselves and also the courts'
selection of the degree of confidence that those judgments must attain in
order to sustain liability in a specific case.
Malone notes that causal judgments, like any other evaluation of
raw evidentiary data, involve interpretation and the drawing of infer-
ences based on the past experience and "judging personality" of the per-
300. Cole, supra note 7, at 778-79; accord A. BECHT & F. MILLER, supra note 8, at 175-76.
301. Cole, supra note 7, at 771-74, 777, 784-91.
302. T. BEAUCHAMP& A. ROSENBERG, supra note 21, at 119-20, 131-37, 145-49; Mackie, supra
note 21, at 24-25, 30-31.
303. H.L.A. HART & T. HONORt, supra note 8, at 45-48; see T. BEAUCHAMP & A.
ROSENBERG, supra note 21, at 141; infra text accompanying notes 366-72. Contra Cole, supra note
7, at 791-92.
1808 CALIFORNIA LAW REVIEW [Vol. 73:1735

son making the judgment. He asserts that, particularly with respect to

the causation issue, "the evaluation which [a person] will make of the
new fact data will necessarily be affected by the purpose he is seeking to
34
serve." 0
Malone's principal argument in support of this assertion is the
apparent effect of purpose on a person's selection of one of many contrib-
uting conditions as "the" cause. As seen in Part I, this argument con-
fuses the tortious-conduct and proximate-cause inquiries with the causal
inquiry. The argument demonstrates only that purpose affects the tor-
tious-conduct and proximate-cause inquiries, not that it affects the causal
inquiry. 0 5
Malone also argues, however, that purpose enters into the causal
inquiry itself. He observes that the process of forming and applying
causal generalizations "is basically one of conjecture, for facts almost
never represent themselves in identical patterns. ' 30 6 Moreover, he notes,
even constant repetition of an identical pattern is not per se sufficient to
support a causal generalization. Something more is needed:
Previous observations by a stargazer may enable him to say with absolute
assurance that when a given constellation rises in the heavens another
constellation will soon come into view over the horizon. Yet, no one in
this modem world would assume that the rising of the first constellation
is a cause of the appearance of the second. The inevitable character of
the sequence as gained from past observations is not enough to establish
the relation of cause and effect. Other portions of our knowledge are too
likely to get in the way of such a conclusion. There must be some accept-
able point of affinity between the new event and old experience that is
satisfactory, and this point of satisfaction can be established only in terms
of the purpose toward which the whole process of decision is being
directed.3 °7
Indeed, something more than mere observation of repeated conjunc-
tion of the two events is needed. The something more is the belief, based
on all our empirically derived knowledge, that the two events are linked
through one or more causal generalizations as cause and consequence.
Malone acknowledges that, regardless of the purpose of the inquiry, "no
one in this modern world would assume that the rising of the first con-
stellation is a cause of the appearance of the second. . . . Other portions
of our knowledge are too likely to get in the way of such a conclusion."
Although our judging and evaluating faculties are involved here, the
inquiry is a factual one in which we judge or evaluate on the basis of
empirically derived causal generalizations that are independent from and

304. Malone, supra note 6, at 61-62.

305. See supra text accompanying notes 17-27.
306. Malone, supra note 6, at 65.
307. Id. (emphasis in original).
1985] CA USA TION IN TORT LAW 1809

not affected by the purpose of the particular causal inquiry.

The second prong of Malone's attack focuses on the problematic
causation cases. He claims that the courts have relaxed the actual-causa-
tion requirement in the overdetermined-causation cases and lowered or
raised the burden of proof in the other types of problematic causation
cases depending on their assessment of the policies underlying the rule
that was violated.30 9
We have already seen that this claim fails with respect to the overde-
termined-causation cases that commentators usually address.3 10
Malone's own discussion of these cases persuasively demonstrates that
the only "relaxation" of the actual-causation requirement that occurs is
the abandonment of the inadequate but-for test in the face of clear evi-
dence of causal contribution.3 1 '
Malone also discusses the "doubtful" causation cases, in which
some safety device or precaution has been omitted that may well have
contributed to the victim's injury, but the plaintiff is unable to establish
that but for the defendant's omission the victim would not have been
injured. Examples include cases in which a corpulent person slips and
falls while hurrying down some unlit stairs, a child drowns in a pool
without a lifeguard, a sailor falls overboard and drowns and the would-be
rescuers are stymied by the lack of proper rescue equipment, or a pedes-
trian slips on an icy sidewalk with an irregular surface. Malone quotes
language from several of these cases that suggests that the courts raised,
lowered, or shifted the burden of proof on causation. He asserts that
their decision to do so depended on the policies underlying the rule that
was violated by the defendant.3 12
But these cases are simply less familiar instances of the overdeter-
mined-causation problem. A full reading of the cases in which the courts
seem to lessen or shift the burden of proof regarding causation usually
discloses that the court was explicitly or implicitly concerned that the
defendant not escape liability under the governing but-for test when the
jury reasonably could conclude from the evidence that it was more likely
than not (the usual burden of proof in tort cases) that the defendant's
omission of the safety precaution contributed to the injury.3 13 Con-

308. Although there are undertones of epistemological skepticism in this portion of Malone's
article, I do not believe that he is arguing that there are no real facts, or that we cannot perceive or
agree on the existence of any facts because of our subjective filtering of sense data. See T.
BEAUCHAMP & A. ROSENBERG, supra note 21, at 16-17, 33-79, 88, 96-97, 112-15, 131-45, 152-55,
259-60; A. BECHT & F. MILLER, supra note 8, at 2-3, 9, 11.
309. Malone, supra note 6, at 67-72, 84-86, 88-89.
310. See supra text accompanying notes 235-60.
311. Malone, supra note 6, at 88-94; accord Carpenter, supra note 262, at 943, 952; see supra
text accompanying notes 241-42.
312. Malone, supra note 6, at 68-79, 85-88, 94-96.
313. See cases cited infra note 335.
1810 CALIFORNIA LAW REVIEW [Vol. 73:1735

versely, the cases which seem to raise the burden of proof are generally
either cases in which the court rigidly adheres to the but-for test despite
good evidence of causal contribution, or cases in which the court believes
the evidence is insufficient to support any well-grounded belief (rather
than pure speculation) even on the issue of causal contribution. 14
For example, it usually is difficult to establish that proper lighting
probably would have prevented the victim's falling down the stairs, or
that proper construction or maintenance of a sidewalk to prevent irregu-
larities in the surface probably would have prevented the victim from
slipping on it when it was covered with ice. But the plaintiff often may
be able to establish in such cases that the lack of light or the irregular
surface more likely than not satisfied the NESS test and therefore con-
tributed to the accident. Thus, the plaintiff may be able to prove that the
lack of light was necessary for the sufficiency of a set of actual antecedent
conditions which included a person of "at least X size" descending the
stairs at a rate of "at least Y steps per minute," and that the sufficiency of
this set was not preempted but rather reinforced by the existence of the
additional conditions of greater size and faster descent. Similarly, the
irregular surface might well have been necessary for the sufficiency of a
set containing a sidewalk of "at least X slipperiness." As Leon Green
clearly recognized, in these cases as well as in the more familiar overde-
termined-causation cases, the but-for test distracts attention from the
31 5
issue of causal contribution.
As long as the but-for test is considered to be the applicable test of
causation in these cases, there will be a strong urge to lessen or shift the
burden of proof when the court perceives that causal contribution may
well exist. While that approach was adopted in the original Restatement
of Torts,3 16 the Restatement (Second) states that the plaintiff bears the
usual (more likely than not) burden of proof in these cases.3 17 Although
the Restatement (Second) deemphasizes but-for language in favor of sub-
stantial-factor language, the but-for language has not been entirely elimi-
nated and the substantial-factor language is typically nebulous.
3 18
Consequently, the overall discussion is hardly enlightening.

314. E.g., Taylor v. City of Yonkers, 105 N.Y. 202, 209, 11 N.E. 642, 644 (1887) (rigid
adherence to but-for test); New York Cent. R.R. v. Grimstad, 264 F. 334 (2d Cir. 1920) (pure
speculation on causal contribution).
315. Green, supra note 8, at 556-59. But see id. at 560-61 (ratifying Malone's argument).
316. RESTATEMENT OF TORTS § 432(1) comment c and accompanying illustrations (1934),
cited in Malone, supra note 6, at 73 n.26.
317. RESTATEMENT (SECOND) OF TORTS § 433B(1) comment b and accompanying illustrations
(1965).
318. See A. BECHT & F. MILLER, supra note 8, at 135-37. The story in England is similar. For
a brief period, the practice was to shift the burden to the defendant when a statutory duty had been
breached to show that the precaution would not have averted the injury. Vyner v. Waldenburg
Bros., 1946 K.B. 50 (C.A. 1945). But then the House of Lords declared that the plaintiff must
1985] CA USA TION IN TORT LAW

The policies or principles being applied in these cases clearly are

those underlying the actual causation requirement itself, rather than
those underlying the particular rule that was violated. Malone's argu-
ments to the contrary are not very persuasive. He argues, for example,
that the shifting of the burden of proof in the hunting-accident cases,
where it is known that one of the negligent defendants shot the plaintiff
but not which one, is based on the very strict attitude toward firearm
accidents, rather than the special causal proof problems in such alterna-
tive-causation cases. 319 He tries to rationalize defendants' greater success
in the missing fire-escape cases than in the man-overboard cases by refer-
ring to the allegedly more restricted reach of the rule in the fire-escape
cases. But his discussion demonstrates that the more restricted reach
occurs not because of any difference in the strength or strictness of the
policies, but because the evidence of actual causation in the fire-escape
cases often shows that the failure to provide the fire escape did not con-
tribute to the injury.32 °
Similarly, he asserts that the burden of proof is not lessened, but
rather increased, in the medical-malpractice cases because of the impor-
tant role fulfilled by doctors who apply "the tenets of an ever-changing
and experimental science. ' ' 32 ' But again his discussion and, even more
so, a note on which he relies322 show that the burden of proof has not
been raised in these cases. Instead, the special difficulties of proof in
medical-malpractice cases, especially in the older cases, 32 3 frequently
have made it difficult for patients to establish actual causation. Expert
testimony by other doctors normally is necessary, and their reluctance
"to do more than express a conjecture on what would have happened
with proper treatment constantly militates against the plaintiff's inter-

"prove his case by the ordinary standard of proof in civil actions: he must make it appear at least
that on a balance of probabilities the breach of duty caused or materiallycontributed to his injury."
Bonnington Castings Ltd. v. Wardlaw, 1956 A.C. 613, 620 (emphasis added) (opinion of Lord Reid).
In subsequent cases the House of Lords has reaffirmed that contribution to the injury is sufficient to
establish cause-in-fact, although the commentators continue to be mesmerized by the but-for tes'
and to view the decisions as exceptions to the actual-causation requirement that were adopted for
moral or policy reasons. E.g., M'Ghee v. National Coal Bd., 1973 Sess. Cas. 37 (H.L. 1972),
discussed in Fraser & Howarth, supra note 227, at 141-42; Weinrib, supra note 7. Weinrib takes a
broader view of actual causation than most, but still argues that policy factors were determinative.
For an earlier case very similar to McGhee, see Gardiner v. Motherwell Mach. & Scrap Co., 1961
Sess. Cas. 1 (H.L.). See supra text accompanying note 261.
319. Malone, supra note 6, at 82-85. The alternative-causation cases are discussed infra text
accompanying notes 342-62.
320. Malone, supra note 6, at 78.
321. Id. at 86.
322. Annot., 13 A.L.R.2d 11, 21-24 (1950), cited in Malone, supra note 6, at 88 n.68.
323. Malone does not mention the conspiracy of silence among medical professionals. This
practice was bolstered by the locality rule which limited testimony to doctors from the same locality
and sometimes even the same school of practice. See, e.g., Brune v. Belinkoff, 354 Mass. 102, 235
N.E.2d 793 (1968); Bowles v. Bourdon, 148 Tex. 1, 219 S.W.2d 779 (1949).
1812 CALIFORNIA LAW REVIEW [Vol. 73:1735

est. '' 324 Usually they will be unable to state that the consequence would
not have occurred but for the defendant doctor's negligent mistreatment
or failure to treat, or even that the mistreatment or failure to treat con-
tributed to the consequence. At best, they often may be able to testify
only that the mistreatment or failure to treat reduced the plaintiff's
chance of recovery-for example, from forty percent to twenty percent,
or from eighty percent to seventy percent.3 25
The problems presented by these reduced-chance cases, as well as
the similar increased-risk and alternative-cause (for example, hunting-
accident) cases, are explored in Part III. For now, it is important to note
that the problems are correctly perceived by courts and commentators as
problems related to proof of causation, which are resolved by turning to
the policies or principles that underlie the actual-causation requirement
itself rather than the policies or principles that underlie the particular
rule that was violated.3 26 Indeed, the few cases that have allowed recov-
ery in the reduced-chance context generally have been medical-malprac-
tice cases, despite Malone's contention that the "policy thrust" of the
327
rules in this area is too "short and timid" to permit such recovery.
Finally, it should be noted that shifting, lessening, or raising the bur-
den of proof on causation does not make the causal inquiry itself any less
factual. The causal inquiry is still an empirical, factual one that results in
a certain degree of confidence that causation did or did not exist in a
particular situation. The decision as to how high that degree of confi-
dence must be, and who bears the burden of producing evidence to sup-
port that degree of confidence, is a distinct issue of policy or principle. In
the criminal law, the prosecution bears the burden of establishing a very
high degree of confidence: "beyond a reasonable doubt." In tort law, the
plaintiff generally bears the burden of establishing a "more likely than
not" degree of confidence. But in each context, the causal inquiry itself is
a factual one.
Nevertheless, a demonstration of frequent ad hoc shifting or lessen-
ing of the burden of proof on causation would indicate that the actual-
causation requirement, rather than being a fundamental criterion of tort
liability, is just one more manipulable doctrine. But no such ad hoc

324. Malone, supra note 6, at 87; see Annot., supra note 322, at 22-23.
325. Malone, supra note 6, at 86-87.
326. See, e.g., sources cited infra notes 335, 339, 348; Delgado, supra note 7; Robinson, supra
note 7.
327. Malone, supra note 6, at 81, 87. See, e.g., Kallenberg v. Beth Israel Hosp., 45 A.D.2d 177,
357 N.Y.S.2d 508 (1974), aff'd mem., 37 N.Y.2d 719, 337 N.E.2d 128, 374 N.Y.S.2d 615 (1975);
Herskovits v. Group Health Coop., 99 Wash. 2d 609, 664 P.2d 474 (1983). For useful reviews of the
medical malpractice cases, see King, Causation, Valuation, and Chance in Personal Injury Torts
Involving Preexisting Conditionsand Future Consequences, 90 YALE L.J. 1353, 1365 n.38, 1367 n.45,
1368-69 n.53 (1981); Note, Increased Risk of Harm: A New Standardfor Sufficiency of Evidence of
Causationin Medical Malpractice Cases, 65 B.U.L. REV. 275, 281-92 (1985).
1985] CAUSATION IN TORT LAW 1813

manipulation has been demonstrated. In almost all the cases that

Malone discusses, the courts have been reacting to the inadequacies of
the but-for test as a test of causal contribution. Very rarely have they
allowed recovery when causal contribution to the injury could not be
established under the more-likely-than-not standard.
The few exceptions-the reduced-chance, increased-risk and alter-
native-causation cases-are cases in which the courts are reluctant to
allow a defendant to escape liability when his tortious conduct may well
have contributed to the injury but it is impossible, given the nature of the
situation, to produce evidence that he actually did contribute. In all
these cases, as Malone admits, liability is never imposed if it is clear that
the tortious aspect of the defendant's conduct did not contribute to the
injury, no matter how strong the policy underlying the rule that was vio-
lated. 2 I argue below, in Part III, that these cases are best understood
as cases that define new legally protected interests, rather than as depar-
tures from the usual strictures of the actual-causation requirement.32 9
In sum, the causal inquiry is indeed a factual, policy-neutral inquiry,
and the actual-causation requirement retains a secure position as a fun-
damental criterion of tort liability.3 3 °

III
CAUSATION, PROBABILITY, AND RISK: PROBLEMS ON THE
FRONTIERS OF TORT LIABILITY

The most serious challenge to the actual-causation requirement has

come from the increasing number of risk-exposure cases. These are cases
in which one or more defendants tortiously exposed the plaintiff to the
risk of an injury that subsequently occurred, but no proof exists that any
particular defendant actually contributed to the injury. Many writers
have asserted that the courts properly ignore or relax the actual-causa-
tion requirement in these risk-exposure cases in order to hold the defend-
ants liable.33
" ' Recently, however, some writers have used the risk-

exposure cases to support a much broader attack on the actual-causation

requirement. Legal economists and others have argued that the tradi-
tional ex post inquiry into actual contribution to the injury should be
replaced by an ex ante probabilistic assessment of increased risk in all

328. Malone, supra note 6, at 71, 77, 78; accord Carpenter, supra note 262, at 943, 947, 952.
329. See infra text accompanying notes 334-62.
330. Interestingly, Malone himself acknowledges these points in his later, less well-known
writings, where he insists that the causal inquiry is "exclusively a fact inquiry" which "should be
maintained utterly devoid of any policy overtones." Malone, supra note 27, at 371; see Note, supra
note 27, at 1540-41 & nn.88 & 89.
331. See Calabresi, supra note 9; sources cited supra notes 6-8.
1814 CALIFORNIA LAW REVIEW [Vol. 73:1735

tort cases.332
I have demonstrated elsewhere that the probabilistic increased-risk
concept cannot be substituted for the actual-causation requirement in the
general run of cases without reaching results that are far removed from
3 33 I
the traditional notions of liability that are applied by the courts.
argue here that, even in the risk-exposure cases, more satisfactory results
are reached by adhering to the actual-causation requirement while recog-
nizing a new type of injury-risk exposure-in certain narrowly circum-
scribed situations. The argument is developed by first analyzing the
more traditional reduced-chance and increased-risk cases and then exam-
ining the currently more topical alternative-causation cases.
Finally, I attempt to clarify the fundamental distinction between ex
ante statements of probability or increased risk and ex post statements of
causal contribution. This distinction helps to clarify the risk-exposure
cases and justify the courts' reluctance to accept naked statistical evi-
dence as proof of causation.

A. The Reduced-Chance and Increased-Risk Cases

As was discussed above, although the plaintiff often cannot prove
that the defendant's tortious conduct was a but-for cause of her injury,
she frequently can prove that it more likely than not contributed to her
injury. 334 In such cases, the courts have held the defendant liable for the
injury. Examples include cases in which the victim slips and falls on
negligently unlit stairs, is washed overboard and there was a negligent
failure to run lines around the boat to prevent such an occurrence, or is
subjected to a negligent delay in treatment of an illness such as bronchial
pneumonia which normally gets worse the longer it is untreated.33 5 Simi-
larly, the courts have recognized that carcinogenic and toxic substances
have cumulative, reinforcing effects, so that a defendant who exposes the
victim to a substance which increases the risk of a certain cancer or ill-
ness is properly considered to have contributed to the subsequent occur-
rence of that cancer or illness, whether or not it might have occurred
33 6
anyway.

332. See Calabresi, supra note 9; Fraser & Howarth, supra note 227; Landes & Posner, supra
note 9; Rizzo, supra note 227; see also G. CHRISTIE, supra note 4, at 246; Shavell, supra note 9.
333. Wright, supra note 9, at 452-55.
334. See supra text accompanying notes 312-27.
335. E.g., Zinnel v. United States Shipping Bd. Emergency Fleet Corp., 10 F.2d 47 (2d Cir.
1925) (missing lifeline); Reynolds v. Texas & Pac. Ry., 37 La. Ann. 694 (1885) (unlit stairway);
Dunham v. Village of Canisteo, 303 N.Y. 498, 104 N.E.2d 872 (1952) (delay in treatment);
RESTATEMENT (SECOND) OF TORTS § 433B(1) comment b and accompanying illustrations (1965);
Green, supra note 8, at 558-59 & n.41.
336. E.g., McAllister v. Workmen's Compensation Appeals Bd., 69 Cal. 2d 408, 445 P.2d 313,
71 Cal. Rptr. 697 (1968); M'Ghee v. National Coal Bd., 1973 Sess. Cas. 37 (H.L. 1972); Clarkson v.
Modem Foundries, [1957] 1 W.L.R. 1210 (Leeds Assizes); see Note, IncreasedRisk of Canceras an
1985] CAUSATION IN TORT LAW 1815

In other cases, however, the plaintiff may be unable to prove that the
tortious aspect of the defendant's conduct more likely than not contrib-
uted to the injury. He may be able to prove only that the tortious
aspect-for example, failure to provide the proper lifesaving equipment
or to properly diagnose and treat the patient-reduced the victim's
chance of avoiding the injury from, say, forty percent to twenty percent.
In these cases, unlike the carcinogenic-exposure cases, there is no incre-
mental, cumulative contribution. The omitted safeguard or treatment
either would or would not have prevented the injury.
One approach to these cases, when the plaintiff cannot establish that
the tortious aspect more likely than not contributed to the injury, is to
reduce the plaintiff's burden of proof to permit recovery even though the
degree of confidence in causal contribution is less than fifty-one percent.
This approach, which has been adopted by a few courts,3 37 weakens the
actual-causation requirement. It does not, however, make the causal
inquiry any less factual.
A second approach is to allow recovery for the reduced chance of
avoiding the injury, rather than for the injury itself. To apply this
approach, the courts would have to recognize the chance of avoiding
injury, in a case where injury actually occurs, as a legally protected inter-
est. If the chance interest is deemed to be legally protected, at least
against "significant impairments, ' 338 a court must then address the
actual causation issue: did the tortious aspect of the defendant's conduct
more likely than not contribute to a (significant) impairment of the
chance interest in this case? If not, the court should not permit the plain-
tiff to recover, no matter how strong the policies underlying the rule that
was violated.
The second approach preserves the integrity of the actual-causation
requirement, while also addressing the basic issues in these cases. Should
the defendant completely escape liability when his tortious conduct
reduced the victim's chance of avoiding an injury that has actually
occurred? This is an issue of policy or principle concerning the types of
interests that should be legally protected. It should be faced and resolved
explicitly, not covertly by diluting the actual-causation requirement.

Actionable Injury, 18 GA. L. REv. 563, 580-81 n.68 (1984). The author of the Note, however,
erroneously assumes that the defendant's contribution was the sole cause if it was independently
sufficient for the occurrence of the cancer. Id.
A defendant who has tortiously contributed to a certain instance of cancer or illness may seek
contribution from other contributing tortfeasors, or from the plaintiff herself if the plaintiff also
contributed (e.g., by smoking). The defendant may even escape liability entirely if he can prove that
the cancer or illness would have occurred anyway as a result of independent nontortious
contributing conditions. See supra text accompanying notes 258-73.
337. See supra note 327.
338. Compare the "significant harm" limitation applicable in private nuisance cases.
RESTATEMENT (SECOND) OF TORTS § 821F (1979).
1816 CALIFORNIA LAW REVIEW [Vol. 73:1735

When the policy issues are carefully distinguished from the causal issue,
the liability analysis is greatly improved. For example, if it is admitted
that the interest being protected in these cases is the chance interest,
recovery should be limited to the value of the lost chance rather than all
the damages attributable to the injury itself. A practical method of valu-
ing the lost chance is to multiply the total damages attributable to the
injury by the amount of reduction in the chance of avoiding it.
The second approach appears to be gaining recognition as the pref-
erable one in these cases.3 39 Under this approach, a plaintiff who is actu-
ally injured can recover for the injury itself if he can prove that the
tortious aspect of the defendant's conduct contributed to the injury. If
he cannot prove this, but can prove that the tortious aspect contributed
to a specific reduction in his chance of avoiding the injury, he can recover
for the reduction in the chance.
Some writers have argued that the courts should go even further.
They would allow recovery for risk-imposition even when no tangible
injury has yet resulted from the imposition of the risk.34 Such a dra-
matic extension of the category of protected interests would raise serious
theoretical and practical problems.3 ' In any event, the issues involved
are not causal ones but questions of policy or principle concerning the
types of interests that are to be protected against invasion. They will be
resolved most clearly if they are seen as such.

B. The Alternative-CausationCases
In the alternative-causation cases, several tortfeasors each indepen-
dently exposed the victim to the risk of a specific injury that subsequently
occurred. It is clear, however, that only one (or a few) of the tortfeasors
actually contributed to the injury. The plaintiff is unable to establish
which one (or few) actually contributed. For example, in the leading
case of Summers v. Tice,342 the plaintiff was struck in the eye by a single

339. See O'Brien v. Stover, 443 F.2d 1013, 1018-19 (8th Cir. 1971); James v. United States, 483
F. Supp. 581, 587 (N.D. Cal. 1980); Herskovits v. Group Health Coop., 99 Wash.2d 609, 622-36,
664 P.2d 474, 480-87 (1983) (Pearson, Stafford, and Utter, JJ., and Williams, C.J., concurring);
PROSSER & KEETON, supra note 7, § 41, at 272; Green, supra note 8, at 558-59; King, supra note
327, at 1376-87; Malone, supra note 6, at 80-81.
340. Note, supra note 336; Note, Increased Risk of Diseasefrom Hazardous Waste: A Proposal
for Judicial Relief 60 WASH. L. REV. 635 (1985).
341. Dworkin, Fear ofDisease andDelayed Manifestationof Injuries: A Solution ora Pandora's
Box?, 53 FORDHAM L. REV. 527 (1984). The courts have yet to accept this argument, even those
courts with the most liberal attitudes toward liability. For a recent example, see Collins v. Eli Lilly
Co., 116 Wis. 2d 166, 191 n.10, 342 N.W.2d 37, 49 n.10, cert. denied, 105 S. Ct. 107 (1984),
discussed infra text accompanying notes 356-57. It should be noted that tangible injury includes
emotional distress, as in the "cancerphobia" cases. E.g., Ferrara v. Galluchio, 5 N.Y.2d 16, 152
N.E.2d 249 (1958).
342. 33 Cal. 2d 80, 199 P.2d 1 (1948).
1985] CA USA TION IN TORT LAW 1817

pellet when the two defendants independently fired their shotguns in his
direction. Obviously, only one of the defendants actually caused the
plaintiff's injury, but there was no way to determine from which gun the
pellet came. The court held the defendants jointly and severally liable for
the injury. It shifted the burden of proof on causation to them, on the
ground that it was better to have the loss fall on the two negligent
defendants than on the innocent plaintiff. 43
Summers represents a significant departure from the usual applica-
tion of the actual-causation requirement. Prosser incorrectly states that
the court "merely extended the rule as to the burden of proof on the issue
of apportionment of damages." 3" The rule that Prosser cites shifts the
burden to the defendants to allocate the loss among themselves only after
the plaintiff has established that each defendant contributed to the
injury.3 45 In Summers, however, we know that one of the defendants is
being held liable for an injury to which he did not contribute. This
knowledge also distinguishes Summers from the reduced-chance cases.
As the number of defendants increases, the departure from the usual
principles of tort liability becomes ever more apparent and difficult to
justify. Thus, if there are N defendants, we know that N - 1 of them are

343. This rationale for joint and several liability was adopted long before Summers in Tidal Oil
Co. v. Pease, 153 Okla. 137, 139-40, 5 P.2d 389, 390-91 (1931) (cattle drank from one or both of two
streams poisoned by different defendants); accord Bowman v. Redding & Co., 449 F.2d 956, 967-68
(D.C. Cir. 1971); Abel v. Eli Lilly & Co., 418 Mich. 311, 325-29, 343 N.W.2d 164, 170-72, cert.
denied, 105 S. Ct. 123 (1984); Cook v. Lewis, [1952] 1 D.L.R. 1 (Can. 1951); RESTATEMENT
(SECOND) OF TORTS § 433B(3) (1965).
Although the Summers court noted that ordinarily defendants have better access to evidence on
causation, this was not true in the particular case, and thus was not the basis for the court's decision.
Sindell v. Abbott Labs., 26 Cal. 3d 588, 601-03, 607 P.2d 924, 929-30, 163 Cal. Rptr. 132, 137-38,
cert. denied, 449 U.S. 912 (1980).
Some writers have tried to justify the imposition of liability in the alternative-causation cases by
arguing that the act of each defendant, in conjunction with the other's act, deprived the plaintiff of a
chance of establishing who caused the injury. E.g., A. BECHT & F. MILLER, supra note 8, at 105
n.155; Weinrib, supra note 7, at 525-26. But this argument proves too much. It could be applied to
any case in which there are alternative potential causes, including those attributable to the plaintiff,
innocent third parties, or natural conditions or events. Acceptance of the argument means that the
defendant is being held liable because he prevented the plaintiff from being consoled by the
knowledge that the defendant was not a cause of his injury. The argument has little merit and
distracts attention from the real issues.
For example, in Haft v. Lone Palm Hotel, 3 Cal. 3d 756, 478 P.2d 465, 91 Cal. Rptr. 745
(1970), the argument is entirely gratuitous. The absence of a lifeguard almost surely contributed to
the drownings, and the court properly insisted that the duty to have a lifeguard remained in effect in
the absence of the statutorily specified warning. The warning, if provided, would simply have
reinstated the assumption-of-risk defense that ordinarily could not be invoked by someone who
failed to comply with the statutory lifeguard requirement. See Osborne v. Salvation Army, 107 F.2d
929, 931-32 (2d Cir. 1939) (no assumption of risk if victim was member of class that statute was
meant to protect); Rovegno v. San Jose Knights of Columbus Hall Ass'n, 108 Cal. App. 591, 291 P.
848 (1930) (absence of lifeguard contributed to drowning).
344. W. PROSSER, supra note 1, § 41, at 243 n.56.
345. See sources cited supra note 264.
1818 CALIFORNIA LAW REVIEW [Vol. 73:1735

being held liable even though they did not contribute to the injury. As N
increases, the probability that any particular defendant is the one who
caused the injury becomes increasingly small. The argument that the
negligent defendants, rather than the innocent plaintiff, should pay for
the injury becomes more a penal argument than a tort argument. As
Professors Harper and James note, "the only fault which should be rele-
vant to the question of civil liability is fault which has caused (or contrib-
uted to causing) the harm."34' 6 The argument even runs counter to a basic
tenet of the criminal law. In effect, the argument asserts that it is better
that N - 1 defendants who did not contribute to the injury should pay for
it than that the one who did cause it should go free. Moreover, the one
who did cause it still may pay for little or none of the damages, due to
contribution or insolvency.
The Summers rule disadvantages plaintiffs as well as defendants.
The underlying rationale of Summers requires that all the alternative
causes of the injury have been tortious and that all the tortfeasors be
joined as defendants. These requirements ensure that the injury was tor-
tiously caused and that the person who actually caused the injury is
before the court. Thus, in a case in which forty persons were firing, but
only seven of them were negligent, the Summers rationale was held inap-
plicable and the plaintiff obtained no compensation.3 47
These problems with the alternative-liability rationale of Summers
have had to be confronted in the recent DES cases. In these cases, the
plaintiff alleges that she contracted cancer as a result of her mother's
ingesting the drug DES while pregnant with her. Typically, the plaintiff
is unable to identify the particular firm or firms who supplied the DES
that her mother ingested twenty or so years earlier. Because there are
hundreds of firms that supplied DES, many of which no longer exist,
most courts recognize that the Summers rationale and requirements
break down in these cases.3 48
In Sindell v. Abbott Laboratories,349 the California Supreme Court
modified the Summers approach to accomodate the most obvious proce-
dural and fairness problems. It stated that the plaintiff need not identify

346. 2 F. HARPER & F. JAMES, supra note 2, § 20.2 n.24 comment, at 95 (Supp. 1968).
347. Burton v. Waller, 502 F.2d 1261 (5th Cir. 1974), cert. denied, 420 U.S. 964 (1975).
348. E.g., Sindell v. Abbott Labs., 26 Cal. 3d 588, 603-04, 607 P.2d 924, 930-31, 163 Cal. Rptr.
132, 138-39, cert. denied, 449 U.S. 912 (1980); Martin v. Abbott Labs., 102 Wash. 2d 581, 591-95,
689 P.2d 368, 375-77 (1984); Collins v. Eli Lilly Co., 116 Wis. 2d 166, 183-84, 342 N.W.2d 37, 45-
46, cert. denied, 105 S.Ct. 107 (1984). But see Abel v. Eli Lilly & Co., 418 Mich. 311, 329-35, 343
N.W.2d 164, 172-75 (approving use of the Summers approach in the DES context), cert. denied, 105
S.Ct. 123 (1984). However, the court, referring to some of the procedural and fairness problems,
reserved judgment on whether modifications should be made to the approach and on the "validity of
any verdict that may result." Abel, 418 Mich. at 331 n.14, 339-40, 343 N.W.2d at 173 n.14, 177
(emphasis in original).
349. 26 Cal. 3d 588, 607 P.2d 924, 163 Cal. Rptr. 132, cert. denied, 449 U.S. 912 (1980).
1985] CAUSATION IN TORT LAW 1819

and join all the firms that were operating in the relevant market at the
time that her mother took the DES, but only enough firms to account for
a "substantial percentage" (undefined but apparently between fifty and
seventy-five percent) of that market. In addition, it replaced the defend-
ants' joint and several liability, which holds each defendant responsible
for the entire injury, with several (separate) liability based on each
defendant's proportionate share of the market.
The substantial-percentage requirement apparently is intended to
ensure that there is a substantial probability that the firm which actually
caused the particular plaintiff's injury is before the court. The limitation
of each defendant's liability to a proportion of the damages equivalent to
its share of the market is intended to ensure that, over a large number of
cases, the defendant is in effect being held liable for the injuries that it
actually caused. That is, the two modifications of the Summers approach
are intended to replicate the actual-causation requirement as applied to
DES injuries in the aggregate, so that each individual plaintiff recovers
for her injury and each defendant firm is held liable, in the aggregate,
only for the injuries that it caused.35 °
However, two aspects of the Sindell approach undermine its ability
to replicate the actual-causation requirement in the aggregate. First, the
Sindell court apparently would allocate one hundred percent of the dam-
ages for each injury among the joined defendants based on their relative
market shares, even though the joined defendants usually would repre-
sent less than one-hundred percent of the market. 351 Thus, the joined
defendants (usually the largest firms) will be held liable for more than
their expected share of the total DES damages. Second, unless the total
number of DES injuries is very large, there will be only a very rough
correlation between the injuries actually caused in the aggregate by a
particular firm and those for which it is held liable under the probabilis-
tic market-share approach. If there is only a relatively small number of
injuries in the aggregate, the extent to which the actual distribution of
injuries among firms varies from the predicted distribution, based on
market share, may be quite substantial.35 2
The Sindell result is much more plausible theoretically if it is viewed
as liability for having tortiously exposed the plaintiff to a certain type of
risk, rather than liability for having caused a certain proportion of the
actual injuries themselves. In effect, the Sindell court recognized a new

350. Id. at 613, 607 P.2d at 938, 163 Cal. Rptr. at 146; see Rosenberg, supra note 7, at 873, 881-
83.
351. 26 Cal. 3d at 612-13, 617, 607 P.2d at 937, 940, 163 Cal. Rptr. at 145, 148.
352. For example, although there is an equal probability of a "heads" or "tails" on each toss of
a coin, the fewer times the coin is tossed, the less likely it is that the actual result of the series of
tosses will be approximately equal numbers of "heads" and "tails." See, e.g., H. BRUNK, AN
INTRODUCTION TO MATHEMATICAL STATISTICS 153 (3d ed. 1975).
1820 CALIFORNIA LAW REVIEW [Vol. 73:1735

legal injury: tortious exposure to a risk that possibly led to a subsequent

injury.3 53 A firm's market share of sales of an identical generic product
such as DES precisely reflects its contribution to the aggregate risk to
which the plaintiff and all others in the relevant market were tortiously
exposed.3 54 If the plaintiff is injured as a result of her exposure to that
aggregate risk, she has a cause of action against each firm which tor-
tiously contributed to the risk.
As in the reduced-chance cases, the damages attributable to such
risk-exposure can be calculated by multiplying the injury damages by the
firm's proportionate contribution to the aggregate risk. Again, if the
plaintiff can prove that a certain firm actually caused her injury by specif-
ically identifying it as the source of the DES that her mother ingested,
she should be able to recover her full damages from that firm. Con-
versely, if a certain firm can prove that it did not provide any of the DES
that her mother ingested, it cannot be held liable for her injury or for
having exposed her to a risk that possibly led to the injury.
Under this risk-contribution approach, it is no longer necessary to
join enough defendants to account for a substantial percentage of the
market, since the focus is on contribution to the risk rather than contri-
bution to the DES injury. For the same reason, the approach avoids the
dissenting judges' concern in Sindell that the defendants are being held
liable without proof of actual causation.3 55 In each case, a defendant
firm is held liable only if the plaintiff proves that the firm tortiously con-
tributed to the aggregate risk that possibly led to the injury.
The risk-contribution approach to the DES cases has been adopted
explicitly by the Wisconsin Supreme Court and implicitly by the Wash-
ington Supreme Court.3 56 Neither the Wisconsin court nor the Washing-
ton court requires that the joined defendants constitute a substantial
percentage of the DES market. However, the Wisconsin court, like
Sindell, holds the joined defendants liable for all the plaintiff's damages.
It apparently allocates the liability severally among the joined defendants
using comparative negligence principles as well as relative market
shares.357 The Washington court initially holds the joined defendants
severally liable for all the damages on a pro rata (equal share) basis.

353. Viewed in this way, the Sindell result is an extension of the risk-exposure approach to the
reduced-chance cases. See supra text accompanying notes 338-39.
354. If there are risk-creating characteristics that vary among the products of different firms,
they should be taken into account. See Rosenberg, supra note 7, at 867 & n.68.
355. Sindell, 26 Cal. 3d at 617, 607 P.2d at 940, 163 Cal. Rptr. at 148 (Richardson, Clark, and
Manuel, JJ., dissenting).
356. Martin v. Abbott Labs., 102 Wash. 2d 581, 594-95, 603-04, 689 P.2d 368, 377, 381-82
(1984); Collins v. Eli Lilly Co., 116 Wis. 2d 166, 191 & n.10, 342 N.W.2d 37,49 & n.10, cert. denied,
105 S. Ct. 107 (1984).
357. Collins, 116 Wis. 2d at 193-200, 342 N.W.2d at 50-53.
19851 CAUSATION IN TORT LAW

However, it allows each defendant that proves its absolute share of the
total market to reduce its liability in accordance with that absolute mar-
ket share. The remaining defendants split the remaining damages.3 58
One other group of alternative-causation cases remains to be dis-
3 5 9 in which an uncon-
cussed. The leading case is Ybarra v. Spangard,
scious patient's shoulder was injured during or immediately after an
appendicitis operation. The California Supreme Court held that the
patient could use the doctrine of res ipsa loquitur against all the doctors
and nurses who had anything to do with him while he was unconscious,
in order to raise an inference of negligent causation against the doctors
and nurses as a group. Each would then be held jointly and severally
liable if he or she failed to come forward with an absolving explana-
tion. 360 The court was clearly determined that a conspiracy of silence
should not deprive the patient of his remedy.
Ybarra seems to go even further than Summers. It is not only
unlikely that all the defendants in Ybarra contributed to the injury, but
also unlikely that they all were negligent. However, the decision can be
reconciled with the tortious-aspect causation requirement if it is viewed
as imposing a form of enterprise liability. The court itself mentioned that
all the defendants could be treated as permanent or temporary employees
of the supervising surgeon or the hospital.3 6 1 When all the defendants
are connected through contractual or commercial relationships into a
common enterprise and can adjust the risks and liabilities among them-
selves, and persons injured by that enterprise ordinarily will have a diffi-
cult time pinpointing the tortious source of the injury, it may be
appropriate to treat the defendants as a group entity-an enterprise-
which tortiously caused the injury, and to let the members of the enter-
prise allocate the liability among themselves or absolve themselves, as
they see fit. The other situations to which the Ybarra rationale has been
applied all fit this analysis.362

C. DistinguishingProbabilityStatements from CausalAttribution

The reduced-chance, increased-risk, and alternative-causation cases
all raise questions about the connection that exists between various types
of probability statements and the actual-causation requirement. These
questions are part of an ongoing debate about the proper role of probabil-

358. Martin, 102 Wash. 2d at 604-06, 689 P.2d at 382-83.

359. 25 Cal. 2d 486, 154 P.2d 687 (1944).
360. Id. at 494, 154 P.2d at 691.
361. Id. at 492, 154 P.2d at 690. This seems to be the preferable approach, since the hospital
and supervising surgeon can control and adjust the risks beforehand and obtain information
afterward more easily than the nurses or orderlies can.
362. See cases cited in PROSSER & KEETON, supra note 7, § 39, at 253 nn. 22, 24 & 25.
1822 CALIFORNIA LAW REVIEW [Vol. 73:1735

istic data and arguments in legal adjudication in general. 363

Due in large part to the lack of an adequate account of causation,
the concept of causal explanation or attribution increasingly has been
confused with predictive statements based either on causal generaliza-
tions or on mere statistical frequency. Some writers have asserted that
the causation requirement itself should be interpreted in probabilistic
terms. Under this notion of "causal linkage" or "probabilistic causa-
tion," the defendant's conduct is considered to have been a cause of the
injury if it increased the risk of occurrence of the injury. 36" Other writers
have not gone so far, but have challenged the courts' reluctance to accept
purely statistical data as evidence of causation or identification. These
writers argue that all evidence, whether it is "particularistic" empirical
data or pure statistics, is ultimately probabilistic, so that the courts' atti-
tude that particularistic data is somehow qualitatively different, more
reliable, or more trustworthy than statistical data is unsupportable. 3 65
These arguments ignore some fundamental distinctions among
causal explanations, causal predictions, and mere statistical reports or
bets. They erroneously assume that because probability judgments
underlie each of these types of statements, that is all that is meant by any
of them. As will be seen, only causal explanations supply the attributive
element that is essential for systems of liability based on individual
responsibility. Mere statements of increased probability, even when
based on causal generalizations, are by themselves insufficient to establish
that attributive element. Probabilities based on causal generalizations,
however, may help to distinguish suggested rival explanations. Pure sta-

363. E.g., L. COHEN, THE PROBABLE AND THE PROVABLE (1977); M. FINKELSTEIN,
QUANTITATIVE METHODS IN LAW (1978); Jaffee, Of Probativity and Probability: Statistics,
Scientific Evidence, and the Calculusof Chance at Trial,46 U. PrrT. L. REV. 925 (1985); Kaye, The
Laws of Probabilityand the Law of the Land, 47 U. CH. L. REV. 34 (1979); Kaye, The Paradox of
the Gatecrasherand Other Stories, 1979 ARIZ. ST. L.J. 101; Nesson, The Evidence or the Event? On
JudicialProofand the Acceptability of Verdicts, 98 HARV. L. REV. 1357, 1377-90 (1985); Rosenberg,
supra note 7, at 855-59, 869-74; Saks & Kidd, Human Information Processing and Adjudication:
Trial by Heuristics, 15 LAW & Soc'Y REV. 123 (1981); Tribe, Trial by Mathematics: Precision and
Ritual in the LegalProcess, 84 HARV. L. REV. 1329 (1971); Williams, The Mathematics of Proof (pt.
1), 1979 CRIM. L. REV. 297; Brilmayer & Kornhauser, Quantitative Methods and Legal Decisions
(Book Review), 46 U. CH. L. REV. 116 (1978).
364. E.g., Calabresi, supra note 9, at 71-72; Fraser & Howarth, supra note 227, at 137-41;
Landes & Posner, supra note 9, at 111-16, 134; Rizzo, supra note 178, at 1009-16, 1037-38; Shavell,
supra note 9, at 468-69; see G. CHRISTIE, supra note 4, at 246; Robinson, supra note 7, at 758-60,
764-65. For my criticism of the views of Calabresi, Landes, Posner, and Shavell, see Wright, supra
note 9.
365. E.g., Kaye, The Laws of Probability and the Law of the Land, 47 U. CHI. L. REV. 34
(1979); Kaye, The Limits of the Preponderance of the Evidence Standard: Justifiably Naked
Statistical Evidence and Multiple Causation, 1982 AM. B. FOUND. RESEARCH J. 487, 488-89, 492
n.22, 514; Rosenberg, supra note 7, at 870-73; Saks & Kidd, supra note 363, at 149-54; Tribe, supra
note 363, at 1344-50, 1361 n.102.
1985] CA USA TION IN TORT LAW 1823

tistical data generally cannot perform even this function and therefore
properly are rejected as evidence of causation or identification.
Causal explanations and causal predictions are applications of
causal generalizations to empirical data. The causal generalizations
themselves are derived inductively from empirical observations of
repeated successions of conditions and events. A causal generalization,
however, is not just a statement of observed statistical correlation. The
essence of a causal generalization is the belief that we attach to the gener-
alization: the belief in its causal or lawlike character. A causal general-
ization asserts that the antecedent conditions produce or cause the
subsequent event-that they are necessary elements of a set of conditions
that is sufficient for the occurrence of the event. It is precisely this qual-
ity of causal generalizations that distinguishes them from mere statistical
reports and gives them explanatory and predictive power.36 6
When we seek to understand and explain the occurrence of an event,
we compare the observed sequence of actual events (the "particularistic"
empirical data) with the antecedent conditions and the associated conse-
quence that are specified in the causal generalization. Usually, very few
of the necessary antecedent conditions must be observed in addition to
the consequence to induce us to believe that the causal generalization is
applicable in the particular circumstances. This is true even when the ex
ante probability of the consequence given these circumstances was very
low-for example, A, a very poor marksman, shoots B from a great dis-
tance with a crooked gun in a high wind under conditions of poor visibil-
ity. 3 67 The empirical observation of the consequence itself is a "vital part
of the evidence which supports an explanation of that event. ' 368
Indeed, as Hart and Honor6 have stated, the force of a particular
causal explanation usually depends on its referring to only a few of the
antecedent conditions and its relying, explicitly or implicitly, on causal
generalizations that are framed in broad or even platitudinous terms. It
is unnecessary, even if it were possible, to explain a particular occurrence
by detailing all the antecedent conditions so that we finally arrive at a
universally true statement of invariable and unconditional causal connec-
tion. As the precision and detail of the description of all the antecedent
conditions increases, our ability to predict the effect improves. Beyond a
certain point, however, the explanatory force of the description does not
improve, but rather lessens as it increasingly becomes a description of a
unique event rather than an instance of some broad generalization appli-

366. T. BEAUCHAMP & A. ROSENBERG, supra note 21, at 131-45.

367. See H.L.A. HART & T. HONORt, supra note 8, at 31-32, 44-48, 487.
368. Mackie, supra note 21, at 36-37.
1824 CALIFORNIA LAW REVIEW [Vol. 73:1735

cable to a multiplicity of different occurrences.

On the other hand, when we use causal generalizations for predic-
tive purposes, we obviously cannot rely on observation of the conse-
quence that is being predicted. We must engage in a much more detailed
investigation and explicit description of the actual antecedent conditions
in order to have any substantial degree of confidence in our prediction. 7 0
We are not able, as when explaining an event that has already occurred,
to ignore all the possibilities concerning what might happen in the light
of the existing conditions and possible future conditions.
It is important to distinguish causal prediction from causal explana-
tion. When we are trying to explain a particular occurrence, the state-
ment that A usually or almost always causes B is a predictive statement,
which is not per se sufficient to persuade us that A caused B in this spe-
cific instance. After all, the "usually" or "almost always" qualifier
admits that there are instances in which A does not cause B, and this
occurrence may well be one of the exception cases. In order to be satis-
fied that the causal generalization ",4 causes B" is applicable in this spe-
cific instance, we must first distinguish any rival explanations that are
suggested by the known or assumed (normal) circumstances. 371
It is at this point that the counterfactual analysis of necessity often
enters into the causal inquiry. Usually the speculative element of the
counterfactual analysis is minimal since only a single antecedent condi-
tion-one of the conditions that constituted the tortious aspect of the
actor's conduct or activity-is changed in setting up the counterfactual
situation, and the alternative explanations, if any, suggested by the
counterfactual situation are generally few and have very different
probabilities. However, when the changed condition generates a number
of possible explanations that have similar probabilities, the speculative,
predictive element becomes dominant and there will be little or no confi-
dence in any particular explanation.3 72
Thus, in the doubtful-causation cases (man overboard, unlit steps,
and so forth) the plaintiff establishes that the tortious condition was a
cause of the victim's injury only if he can convince the trier of fact that
his suggested causal explanation is more plausible than any suggested
alternative causal explanations. If he is only able to prove that the tor-
tious condition reduced the chance of avoiding the injury (increased the
risk of its occurrence) but cannot sufficiently distinguish possible alterna-

369. H.L.A. HART & T. HONORt, supra note 8, at 44-47; see T. BEAUCHAMP & A.
ROSENBERG, supra note 21, at 141, 295-99, 301-02.
370. H.L.A. HART & T. HONORS, supra note 8, at 47; Mackie, supra note 21, at 37.
371. H.L.A. HART & T. HONORt, supra note 8, at 31-32, 48-49. See generally L. COHEN, supra
note 363.
372. See supra text accompanying notes 296-303.
1985] CA USA TION IN TORT LAW 1825

tive explanations of how it actually occurred, he has established only that

the tortious condition contributed to the risk, not to the actual occur-
rence of the injury. He therefore should be able to recover from the
defendant only if exposure to the risk is itself considered37to 3
be a legal
injury, and his damages should be measured accordingly.
In the alternative-causation cases, the problem is significantly differ-
ent. The Summers problem does not arise simply because each defendant
is equally likely (or unlikely) to have been the one that fired the shot that
wounded the plaintiff. It arises because we are trying to use pure or
"naked" statistical probabilities, not based on any causal generalization,
to determine what actually happened. Thus, the problem exists in the
DES cases even though one defendant may have controlled more than
fifty percent of the market and is therefore most likely, from a statistical
standpoint, to be the defendant who caused any particular injury.3 74
Naked statistical data generates very little confidence in deciding who
actually caused the injury. For example, few people would be willing to
say that Joe caused Sue's death simply because Joe fired seven bullets in
her direction while Mike only fired three. No one would treat Joe as the
cause, no matter how many more bullets he fired than Mike, if the bullet
that killed her were identified by its markings as one that came from
Mike's gun.
As evidence fitting a certain causal generalization builds up, it gen-
erates an increasingly strong belief in the existence of the asserted causal
connection.3 75 The markings on the bullet are "particularistic" empirical
data that fit into a causal generalization and therefore have explanatory
force. On the other hand, naked statistical reports of who fired the most
shots in Sue's direction do not fit into any causal generalization and thus
do not generate any such belief.376 They are useful for placing a bet on
373. See supra text accompanying notes 312-29 & 334-41.
374. Eli Lilly & Co. may well have been in this position. Levine, Gilding the Lilly, TRIAL, Dec.
1984, at 19-20. The Sindell court carefully noted that an inference of causation by any one DES
manufacturer based on probability would have failed "if we measure[d] the chance that any one of
the defendants supplied the injury-causing drug by the number of possible tortfeasors." Sindell 26
Cal. 3d at 603, 607 P.2d at 931, 163 Cal. Rptr. at 139 (emphasis added); see also id. at 611-12, 607
P.2d at 936-37, 163 Cal. Rptr. at 144-45. Implicitly, a probabilistic "more likely than not" inference
would have been possible against Eli Lilly & Co. if the chance were measured by market shares
rather than by the number of defendants, but, contrary to David Kaye's suggestion, this was not
what the court was after. See Kaye, The Limits of the Preponderanceof the Evidence Standard:
Justifiably Naked Statistical Evidence and Multiple Causation, 1982 AM. B. FOUND. RESEARCH J.
487, 508, 514. Rather, the court was unwilling to hold a defendant liable for the entire damage
based simply on naked statistical inference. Kaye's article misses a critical element when it treats
proof as a simple matter of probabilities and ignores the need to induce the belief that a causal
relationship actually existed.
375. L.J. COHEN, supra note 363, passim.
376. It is this distinction that explains the difference between Judith Jarvis Thomson's
"external" evidence (naked statistical report) and her "internal" evidence (particularistic evidence
that fits into a causal generalization). Thomson, Remarks on Causation and Liability, 13 PHIL. &
1826 CALIFORNIA LAW REVIEW [Vol. 73:1735

who caused Sue's death, but they do not help us to determine whether
Joe or Mike actually was the cause.3 77 At most, they tell us the relative
amounts of ex ante risk that Joe and Mike imposed on Sue, when they
are combined with the causal generalization that links firing a gun in
someone's direction with killing that person. Thus, they can be used to
allocate liability that is based on exposing someone to the risk that was
realized. Nevertheless, they are of little use in establishing who actually
caused Sue's death.
The courts generally have perceived the critical distinction between
naked statistical evidence and causally relevant particularistic evidence.
For example, in Smith v. Rapid Transit, Inc.,3 78 the court refused to
allow the jury to infer that one of the defendant's buses caused a particu-
lar accident when the evidence merely showed that the defendant's buses
were the most frequent users of the route on which the accident
occurred. Unless the statistical evidence is so compelling that it pre-
cludes any possible alternative explanation, it will be insufficient by itself
to induce the belief that the asserted causal connection actually existed.
In sum, so long as tort liability continues to be based on individual
responsibility, liability will be imposed on a defendant only if it is
believed that the tortious aspect of his conduct actually contributed to
the specified legal injury. This belief is the essence of a causal explana-
tion, as distinguished from mere probabilistic statements of increased
risk. The belief will arise in a particular case only if there is sufficient
evidence that not only supports the suggested causal explanation involv-
ing the defendant's tortious conduct, but also makes it the most plausible
suggested explanation. Ordinarily, only particularistic evidence fitting
the relevant causal generalizations can accomplish this task.

CONCLUSION

Although causation intuitively seems to be a rather simple concept,

it is in fact a complex and subtle one that long has resisted efforts to
articulate a precise definition. The repeated failure of such efforts during

PUB. AFF. 101, 127-33 (1984); see also Nesson, supra note 363, at 1377-79, 1383-85 (acceptable-
stories theory).
377. Many writers have confused the betting odds that a person is willing to accept on the
existence of a certain fact with the belief that the person actually has in the existence of that fact.
E.g., Kaye, The Paradox of the Gatecrasherand OtherStories, 1979 ARIZ. ST. L.J. 101, 105; Tribe,
supra note 363, at 1346-48; Wagner, Book Review, 1979 DUKE L.J. 1071, 1072-73 & n.6. As the
text indicates, a willingness to accept betting odds does not necessarily imply any belief in the actual
existence of the fact in question. Jonathan Cohen has written extensively and persuasively on the
inductive nature of causal belief and the inapplicability of mathematical probability theorems to such
belief. L. COHEN, supra note 363, summarized in Schum, A Review of a Case Against Blaise Pascal
and His Heirs, 77 MICH. L. REV. 446 (1979); see T. BEAUCHAMP & A. ROSENBERG, supra note 21,
at 312-14; Nesson, supra note 363, at 1385-90.
378. 317 Mass. 469, 58 N.E.2d 754 (1945).
1985] CAUSATION IN TORT LAW 1827

this century has led to frequent confusion of the causal and noncausal
issues in tort scholarship. More recently, it has led to a general denigra-
tion of the significance and even of the meaning of the concept of causa-
tion in tort law.
In this Article, I have argued that the concept of causation has a
definite substantive content that was first elaborated by the philosophers
David Hume and John Stuart Mill. I have tried to demonstrate the ana-
lytical clarity and power that results from a proper understanding of the
concept of causation and its connection, through the tortious-aspect cau-
sation requirement, to the concept of legal responsibility. A voluminous
literature that confuses the two concepts can be sorted out and inte-
grated. Problematic cases can be solved. Restrictions on naked statisti-
cal evidence can be explained and justified. The policy issues posed by
recent cases on the frontiers of tort liability can be correctly identified.
I have concentrated on these descriptive issues and problems. How-
ever, this Article also has important normative implications. The deni-
gration of the concept of causation during this century has seriously
eroded the traditional view of tort liability as a system of corrective jus-
tice, which protects individual autonomy by redressing injuries if and
only if they were caused by the tortious aspect of the defendant's con-
duct. Tort scholarship instead has been dominated by books and articles
that view tort liability as a (poor) system for providing general compen-
sation for losses 379 or as a (poor or great) system for maximizing social
wealth. 8 0
But the concept of corrective justice embodied in the tortious-aspect
causation requirement clearly continues to control the decision of actual
tort cases. Despite more than a half-century of academic assaults on the
causation requirement, the courts almost invariably reach results that are
consistent with it. The requirement and the underlying concept of cor-
rective justice exert a very strong normative pull, even when judges do
not articulate that pull but rather couch their decisions in the currently
fashionable language of loss-spreading and wealth-maximization.
As I have noted elsewhere, 3 1 the concept of corrective justice has
fallen on hard times in recent years because of difficulties encountered in
defining its two principal elements, causation and rights, rather than
from any normative or descriptive weakness in the concept itself, which
continues to be both more normatively appealing and more descriptively

379. E.g., R. KEETON & J. O'CONNELL, BASIc PROTECTION FOR THE TRAFFIC VICTIM 1-3, 11-
75, 249-50, 261-65 (1965); Englard, supra note 11, at 28-29, 62, 67-69. See generally Symposium:
Alternative Compensation Schemes and Tort Theory, 73 CALIF. L. REV. 548 (1985).
380. E.g., G. CALABRESI, supra note 100, at 239-87 (poor system); Landes & Posner, The
Positive Economic Theory of Tort Law, 15 GA. L. REV. 851 (1981) (great system).
381. Wright, supra note 9, at 435-36, 455-56.
1828 CALIFORNIA LAW REVIEW [Vol. 73:1735

accurate than any alternative theory of liability. This Article has defined
and elaborated the causal element in the concept of corrective justice. In
another article, now in progress, I address the rights element by elaborat-
ing a principled theory of tortious conduct. I hope, through both arti-
cles, to make explicit the principles that have provided the foundation for
tort liability over the centuries, but which require more detailed elabora-
tion in these skeptical times.