(This article was published in the Boston University Law Review, May-July 1986)
Summary
... My thesis, broadly stated, is that "probability" as we use the term in law, particularly in
the civil standard of proof, is not a hard-edged mathematical concept. ... An imaginative
scientist exploring the hypothesis that a given toxic agent causes cancer is very likely to
suspend scientific judgment on the ultimate question of causation until more testing or study
can be done to eliminate alternative hypotheses. ... Dr. Carnow based his opinion on data
showing exposure, on a suggestive pattern of symptoms and timing, and on a wealth of
information drawn from the scientific literature showing that animals and humans exposed to
dioxin contracted cancer. ... Since that is just the kind of question statistical science
addresses, it is little wonder that Judge Weinstein thought the question should be answered
according to the methodology and standard of proof science would consider appropriate --
proof by epidemiological study. ... But if a jury could rationally reach that conclusion for Lilley,
must it not also follow that a jury could rationally conclude that dioxin caused cancer among
the population of soldiers exposed in Vietnam? Were Judge Weinstein to allow a medical
diagnostician's opinion to satisfy Lilley's burden of proving causation, he could not refuse to
allow a similar expert opinion to satisfy the burden of proof for the class, regardless of what
epidemiology had to say. ...
I. INTRODUCTION
For this conference on Probability and Inference in the Law of Evidence I have chosen to
analyze a specific case. This might at first seem odd for a conference cast at a very high level
of abstraction, yet the case brings into focus a central problem about proof and the nature of
inference in judicial dispute resolution. It aptly highlights the relationship between what a
litigant must prove at trial, and the standards of proof and rules of evidence applied in proving
it.
My thesis, broadly stated, is that "probability" as we use the term in law, particularly in the
civil standard of proof, is not a hard-edged mathematical concept. It is, rather, a concept that
incorporates less rigid ideas of justice and reflects the judicial function of resolving disputes in
the real world, where values shift and knowledge is uncertain. An outcome is "probable" if it
best accomplishes a just and acceptable resolution of the dispute. Probability, as a legal
concept in the law of proof, suggests wisdom, probity, and approbation -- not favorable betting
odds. Note 1
A cornerstone of traditional judicial dispute resolution is the capacity of courts to produce an
acceptable conclusion about what happened on the occasion of a past disputed event. A
conclusion about the facts in a disputed event is the necessary requisite for courts to apply
legal rules in the form, "Given that such and such happened, the results that follow shall be . .
. ." A court moves from conclusions about the evidence to a conclusion about the event by
applying the standard of proof. In many cases the transition is easy. For example, the typical
negligence case pits conflicting witnesses to an event against each other and permits the jury
to conclude whether D's [*522] negligent injury of P probably happened. On that basis, the
court makes the transition to an operative legal conclusion that D negligently injured P, and
proceeds accordingly. The conclusion about the event provides a history of what happened
that serves as an acceptable predicate for asserting the legal rule and imposing a legal
sanction.
In other cases, however, the transition from evidence to event is not so easy. In the famous
blue bus hypothetical, Note 2 for example, one can conclude from the evidence that the plaintiff
was probably injured by a blue bus, and yet not be entirely satisfied with the conclusion that a
blue bus injured the plaintiff. Note 3 The evidence seems not to be about the specific event in
question, but rather about a class of events (all cases involving unidentified buses). It suggests
only that the defendant causes a majority but not all of a class of events. It offers no basis for
a leap to a conclusion about any specific event. Although the information about the class
suggests it is more probable than not that the defendant caused any given accident, a finding
for any one plaintiff would logically extend to every plaintiff in the class.
The blue bus and hypotheticals like it present cases in which we know we do not have a
conclusion about what happened with respect to a specific event. Although we can draw
logical conclusions about a class of events that includes the event in question, no mode of
evaluation or deliberation is available to rationalize a conclusion about the specific event. In
effect, despite statistical probability, a determination about what happened seems speculative.
Probability alone does not provide a satisfying basis for accepting and deferring to a finding
that the defendant injured the plaintiff in the given case. This suggests that the acceptability of
a conclusion is not a simple function of mathematical probability, but rather is a complex
matter of communication that depends on the nature of the issue, the process of decision, and
the purposes and audiences the conclusion serves. Viewed [*523] broadly, the standard of
proof consists not only of the guidance courts give juries, but also of the evidentiary rules of
admissibility courts use to shape the body of evidence the jury considers, and the rules of
sufficiency by which courts decide whether juries will be allowed to consider the evidence.
Conceptualizing these rules in terms of objectively determinable probability Note 4 threatens to
sever them from their underlying purpose.
The case I will discuss, a part of the Agent Orange Litigation, illustrates this tension. The case
turns on the nature of the basis for expert testimony, and suggests different approaches to the
concept of probability that are appropriate to scientific statistical proof on one hand and legal
proof on the other. The case turns on allegations that the United States Government and
various chemical companies negligently exposed soldiers in Vietnam to dioxin, thereby
causing them to suffer multiple serious afflictions including cancer. Research scientists have
not conclusively demonstrated that dioxin is a human carcinogen, but they consider exposure
to the substance a great risk. What type of evidence, if any, will satisfy the plaintiff's burden of
proving causation? Should a medical diagnostician be allowed to give his opinion that dioxin
caused a particular plaintiff's cancer, even though his opinion is based on a proposition to
which scientists have not yet subscribed? Should a court permit a verdict for the plaintiff
predicated on the testimony of a medical diagnostician when statistical scientists say they
can't tell? Defendants contend that standards of proof that would permit such testimony and
support a verdict are unscientific. Plaintiffs respond that the legal standards of proof must be
more liberal than scientific standards because courts must seek justice as well as ultimate,
unassailable, scientific truth. Plaintiffs argue that courts, lacking absolute certainty and full
understanding, must aim at a just approximation of truth. This conflict between law as science
and law as justice can be understood by students of evidence as a dispute about what we
mean by "probable" in law.
John Lilley was a so-called "opt-out" Note 6 plaintiff in the Agent Orange Litigation. Lilley's job in
Vietnam was to empty large drums of Agent [*524] Orange into the tanks on C-123's. Under
field conditions, the chemical was constantly spilling. According to a fellow serviceman who
worked with him, Lilley practically bathed in the stuff. Note 7
Shortly thereafter, Lilley developed
skin lesions, boils, and skin discoloration called chlorachne, which is a virtual signature of
dioxin poisoning. Lilley also developed neurological and immunological problems including
numbness in his hands, a persistent cough, and a sore throat so severe that he required an
adult tonsillectomy.
After Lilley returned from Vietnam, his wife became pregnant; the baby was stillborn at 5
months gestation. Mrs. Lilley became pregnant again and carried this child to term; the baby
was born with a lung disorder and an abnormal susceptibility to infectious diseases. In 1970,
at age 40, Lilley was diagnosed as having lymphatic cancer and told he had six months to
live. His spleen became diseased and was removed. He had a heart attack in 1973. His
cancer progressed to leukemia. After a second heart attack in 1975, Lilley died of leukemia in
1976.
Our government's use of Agent Orange during the Vietnam War was, according to one view,
blatant chemical warfare: we sprayed tons of "defoliant" from giant aircraft onto the enemy's
supply lines. Our government did not admit that these defoliants, which contained dioxin, were
highly toxic to humans. Note 8 Doing so undoubtedly would have heightened domestic resistance
to the program, and further fueled anti-war resistance. Our government masked the defoliants'
toxic nature by calling them innocuous color names: White, Blue, Orange. The soldiers who
worked with Agent Orange were neither advised of its nature nor warned to take precautions in
handling it. Agent Orange containers carried no warning labels, no skull and crossbones.
Soldiers were neither given special protective clothing, nor told to shower and change clothing
after exposure. The safety of our servicemen was comprised not to gain a battlefield advantage
over the enemy, but rather to stave off criticism back home of the defoliation program.
[*525] The Agent Orange Litigation took two paths. Many thousands of veterans who claimed
to have been injured by Agent Orange joined a class action. SUP>Note 9 Approximately three hundred
plaintiffs were "opt-outs" who refused to join the class, preferring to prosecute their claims
individually. Note 10 All the Agent Orange cases were consolidated, and ultimately came before
Judge Weinstein in the United States District Court for the Eastern District of New York. SUP>Note 11
The class action plaintiffs accepted a $180 million settlement, which Judge Weinstein
approved. Note 12 The opt-out plaintiffs, including Lilley, received nothing because Judge
Weinstein granted the defendants' motions for summary judgment. In Judge Weinstein's view,
the epidemiological studies on Agent Orange so clearly failed to prove any causative link
between exposure to Agent Orange and cancers, birth defects, and other illnesses, that
neither Lilley nor any of the other opt-out plaintiffs presented a factual dispute. SUP>Note 13 The
veterans in the class action, according to Judge Weinstein, escaped summary judgment only
because they had settled. Note 14 The propriety of Judge Weinstein's approval of the settlement
and his grant of summary judgment is now on appeal before the United States Court of
Appeals for the Second Circuit.
Considering the magnitude and complexity of the Agent Orange Litigation, Judge Weinstein
achieved a remarkable outcome, truly a tremendous accomplishment. He pursued a forceful,
practical approach, that produced the largest settlement on record. Note 15 His ruling against the
opt-out plaintiffs [*526] undoubtedly reflected his view that a class-wide approach was both
appropriate and necessary to resolve the overall controversy. From Judge Weinstein's
perspective, he gave the opt-outs fair warning by urging them to join the class, in effect telling
them they would get nothing unless they did. Note 16 Thus the overall result, if it stands on
appeal, promises compensation for a mass of plaintiffs, and provides at least some sense that
justice is being done.
Yet Judge Weinstein built his resolution of the controversy on a conception of proof that
seems wrong to me. His view reflects an erroneous and hard-edged statistical concept of
probability, which obscures the difference between law and science. Although he has settled
the Agent Orange cases, his approach, if it becomes dogma, will gravely incapacitate the
disputeresolving powers of courts in toxic tort cases. It leaves entirely to defendants the range
between what jurors and members of the public can rationally believe, and what statisticians
can prove.
The Lilley decision poses the question: what basis must an expert have to render an opinion
on causation sufficient to satisfy a plaintiff's burden of proving that his injury was caused by
exposure to a toxic chemical? The etiology of cancer is not yet known; scientists cannot
explain at a cellular level the mechanism of causation. Animal studies show that dioxin
causes cancer in animals, but there is no epidemiological study that shows statistically that
dioxin causes cancer in humans. Should this preclude a doctor from opining, based on the
scientific and medical literature relating dioxin exposure to disease in animals and humans,
and on his knowledge of the history and symptoms of an individual plaintiff, that dioxin caused
the injury?
III. CLINICAL PROOF OF CAUSATION: THE STANDARD OF THE TREATING DOCTOR
Lilley's case on causation rested primarily on the testimony of an expert witness, Dr. Carnow,
whose affidavit Judge Weinstein considered on summary judgment. Note 17 Dr. Carnow assumed
that Lilley had been heavily exposed [*527] to dioxin in Vietnam,Note 18 and that subsequently
he had not been further exposed to carcinogens.Note 19 Proof of these facts would have been
offered at trial through other witnesses, including Lilley's fellow servicemen and his wife. Dr.
Carnow had reviewed Lilley's medical history, the nature and extent of his exposure to dioxin,
and the timing of the exposure relative to the appearance of Lilley's afflictions. He then
concluded "to a reasonable degree of medical certainty" that Lilley's cancer, cardiac, and
neurological problems were caused by dioxin poisoning in Vietnam.Note20
Dr. Carnow's affidavit presented his clinical judgment. His diagnosis reflected a medical
judgment based on his knowledge of the scientific research done on the physical effects of
exposure to dioxin, combined with the intensity of Lilley's actual exposure to the substance,
and the pattern of Lilley's medical symptoms. Although medical experts reason this way in
almost all lawsuits involving the cause of a litigant's illness, Judge Weinstein rejected Dr.
Carnow's opinion in this case as "unfounded."Note 21 As Judge Weinstein saw it, causation
could not be established based on the available information because: "No acceptable study to
date of Vietnam veterans and their families concludes that there is a causal connection
between exposure to Agent Orange and the serious adverse health effects claimed by
plaintiff."Note 22
This is not the traditional standard for admitting an expert medical opinion. That standard
requires the expert to be qualified, to have used respected methodology in arriving at his
opinion, and to express confidence in his opinion to a reasonable degree of medical certainty.
The doctor's opinion must be such that a layperson could accept the opinion based on his
expertise as an authority in the field. These evidentiary requirements seek to incorporate the
standard that the expert uses in his own profession.
A hypothetical illustrates the nature of the opinion a diagnosing doctor can customarily offer
on legal causation. Suppose hypothetically that Dr. Carnow were treating Lilley for his cancer
and had two options. Option 1 is to treat Lilley with Drug X. If the cancer was caused by
dioxin, Drug X would cure Lilley; if dioxin was not the cause, Drug X would kill him. Option 2 is
to treat Lilley with Drug Y. If Lilley's cancer was not caused by dioxin, then Drug Y would cure
Lilley; if dioxin was the cause, Drug Y would kill him. Assume further that there are no other
treatments, and that without treatment Lilley's condition is hopeless.
[*528] This hypothetical poses a treatment decision for a doctor -- not an abstract issue of
scientific proof. If Dr. Carnow would choose treatment X, and would do so with a reasonable
degree of confidence in his choice based on his medical expertise, then he would be treating
his patient according to his best medical judgment, and would hold his opinion on causation to
a "reasonable degree of medical certainty."Note 23
This standard of the treating doctor is the typical juridical standard of proof necessary for the
doctor's opinion to serve as a basis for a jury's conclusion about what happened.Note 24 To be
sure, the doctor is not saying he is [*529] absolutely or scientifically certain of the diagnosis,
but to expect him to be so would discourage treatment in most medical situations. Certainty
is a false god here, as elsewhere in judicial proof. A qualified medical diagnostician is familiar
with the scientific and medical literature. He assesses the significance of experiments and
studies, not in the technical scientific sense of the statistician, but in an intuitive way.Note 25 He
anticipates what the scientist would be able to prove if he could structure the perfect study,
the perfect experiment. Lacking complete information, the diagnostician gives his best
judgment. By its nature this judgment is not, of itself, scientific proof, but it may nonetheless
constitute legal proof.
The legal standard generally requires the expert to hold and express his opinion with
assurance. Mere possibility is not enough, and complete assurance is not necessary. The
legal standard requires neither absolute certainty, nor the exacting level of certainty that
scientists employ for the demonstration of scientific propositions. There is a difference, then,
between the testifying expert and the lab scientist.Note 26 An imaginative scientist exploring the
hypothesis that a given toxic agent causes cancer is very likely to suspend scientific
judgment on the ultimate question of causation until more testing or study can be done to
eliminate alternative hypotheses. A doctor or lawyer or judge, on the other hand, often does
not have the luxury of postponing a decision. Often he must make a judgment once he
reaches a reasonable working conclusion.Note 27 Likewise, we ask juries to come to conclusions
without [*530] insisting on or waiting for scientific demonstration. The legal standard of proof
would seem to require only a rational basis for the expert's opinion -- a standard far short of
scientific demonstration.Note 28
IV. THE RELATION OF CLINICAL AND STATISTICAL PROOF
There is, of course, a rational linkage between scientific and individual clinical diagnosis. How
can a doctor know that given symptoms result from a particular cause in the absence of
empirical data? Since causation cannot be directly observed in a case of cancer (contrast a
statement that bleeding was caused by a cut at the location where the blood appeared), how
can the doctor's clinical judgment, to the extent that it goes beyond what has been
scientifically proven, be based on rational inference? At best, the doctor has observed or heard
about an unscientific sample of patients, or is extrapolating from the data reported in scientific
studies. One can easily imagine the sample the doctor has observed being so limited and
disparate that a trial judge could consider the doctor's opinion speculative and hence
inadmissible. There is something, then, to the structure of Judge Weinstein's approach. If
many carefully controlled studies showed that eating carrots had no conceivable connection to
cancer in humans or animals regardless of dose, a judge would certainly be justified in
granting summary judgment or directing a verdict in favor of a defendant carrot company,
notwithstanding a medical diagnostician's opinion that the cancer was caused by plaintiff's
carrots.Note 29
[*531] This extreme example emphasizes the distinction between the justifiable power of a
judge to foreclose rank speculation from the unjustified exclusion of a medical expert's opinion
in the Lilley case. It also emphasizes the importance of defining the standards that are
appropriate in judging statistical and clinical evidence. Judges face a delicate task in
assessing the adequacy of the basis for an expert's opinion. At a basic level, the legal system
uses the idea of expertness to accomplish its objective of reaching a conclusion by
subsuming the very question of whether a conclusion is possible within the domain of
expertise. Thus, if a sufficiently credentialed expert says he has reached a conclusion to a
reasonable degree of certainty, that is generally enough. The very question of whether it is
possible to come to a conclusion to a reasonable degree of certainty is considered a matter
the court is not sufficiently competent to judge. Judges, after all, are lay persons, no matter
how well-read they are in science. A judge normally disqualifies an expert's opinion only when
the basis for that opinion is inadequate even as judged by a layperson. When a judge goes
further and disqualifies an opinion that is based on a wealth of data, he is almost certainly
intruding on the jury's role of assessing the weight to be given the opinion because he will
necessarily be assessing the significance of the data -- a task for which he is not qualified,
and for which the expert witness is. This was Judge Weinstein's mistake.
Dr. Carnow based his opinion on data showing exposure, on a suggestive pattern of
symptoms and timing, and on a wealth of information drawn from the scientific literature
showing that animals and humans exposed to dioxin contracted cancer.Note 30 The scientific
literature contains more than one hundred epidemiological studies of the effects of dioxin on
animals and on humans (resulting from industrial accidents). Judge Weinstein said these
studies were useless in establishing causation for the plaintiffs.Note 31 He discounted one
[*532] study on which Dr. Carnow relied because the scientists who conducted it concluded,
as good scientists will, that more research needed be done to determine the mechanism of
dioxin's action before a scientific conclusion that it causes cancer could be drawn.Note 32 He
discounted another study, which showed that dioxin poisoning resulted in cancer in animals,
because it cautioned against extrapolating from the experimental animals to humans.Note 33
Although the research scientists who performed these studies were not yet satisfied to a
scientific standard that dioxin causes cancer in humans, their data and results are surely
evidence for doctors who must make clinical diagnoses. Lab rats are extremely healthy, bred
to be strong, without a history of disease or defect. Lilley, by constrast, was a typical adult
male in the Armed Forces, a smoker, and an airplane engine mechanic (and thus prone to
benzene exposure). If a chemical shows carcinogenic effects in a population of healthy lab
animals, a doctor might anticipate similar results among an equivalently exposed population of
ordinary, relatively unhealthy human beings.Note 34 Once Judge Weinstein rejected Dr. Carnow's
opinion, he had to assess for himself the significance of the scientific data. Indeed, he
ultimately opposed his own expert opinion to Dr. Carnow's: "John Lilley's long career 'tearing
engines apart' makes it far more likely that exposure to benzene or radiation from
contaminated aircraft caused his lymphosarcoma."Note 35
V. VULNERABILITY TO THE RISK OF BEING PROVEN WRONG
Two considerations might be thought to make medical diagnostic methodology inappropriate
as a basis for a verdict in Lilley's case. Both relate to limitations on the capacity of the judicial
process to project an acceptable and durable account of what happened.
First, the degree to which a verdict would be vulnerable to being proven wrong may affect the
standards of proof, including the standard for judging the sufficiency of expert testimony to
support a verdict. To see this, distinguish [*533] two cases. In Case A, a typical tort case,
perhaps an intersection accident, there is a dispute about what happened on the occasion of a
past event. The dispute results from the absence of an authoritative witness to the event, or
perhaps from conflicts among several witnesses. If a judicial factfinder could somehow watch a
replay of the disputed event, he could definitively resolve it. Factfinders have no such power,
and therefore cannot determine what happened with absolute certainty. They can, however,
evaluate evidence and arrive at the most probable conclusion. Assume that the factfinder, after
evaluating all the evidence, considers it 70% likely that the event occurred as the plaintiff
alleges. The result will be a verdict for the plaintiff.
Case B, a case like Lilley's, involves a determination beyond the current boundary of science
that implicates an ongoing, disputed supposition about how nature works. There is a dispute
about whether dioxin caused the plaintiff's injuries, but resolution involves a conclusion about
something not observable or otherwise definitively knowable at our current level of knowledge.
Even with the aid of a time machine to watch the event unfold, the factfinder could not
definitively say what happened. Assume the factfinder, after evaluating all relevant evidence
including expert testimony, considers it probable (say 70% likely, just as in Case A) that
dioxin caused the plaintiff's cancer (and has no problems with any other part of the plaintiff's
case), with a resulting verdict for plaintiff.
The verdicts in Cases A and B, though resting on equally probable conclusions about what
happened (as subjectively assessed by the factfinder), differ in their vulnerability to being
proven wrong. In Case A, the more time that passes between the event and the trial, the less
likely the verdict will subsequently be proven wrong. New evidence may be discovered, but
none is being generated. By contrast, the decisive issue in Case B is one in which more
evidence will be generated as time goes by. Presumably science will eventually figure out just
how cancer is caused, and determine to a scientific certainty whether dioxin is a human
carcinogen. Since the factfinder's probability assessment is a prediction about what the future
will show, the verdict in Case B is likely to be tested, and may very well prove to be wrong.
Case A is thus more suited to judicial resolution than is Case B. Imagine, for example, that a
study will be completed in two years that, all agree, will definitively determine the effects of
dioxin in humans. I suspect that any court would hesitate to allow judgment for Lilley without
waiting for the definitive test. Judgment against Lilley on grounds that he had not met his
burden of proof would leave the court less exposed to being proven wrong than would a verdict
in Lilley's favor. The definitive test, if it favored Lilley, might embarrass his counsel for pressing
ahead too soon, but it would not bother the court.
I do not mean to suggest that a risk of being proven wrong should preclude courts from
rendering affirmative verdicts. Preoccupation with such risk would incapacitate the judiciary
from dealing with some of our society's [*534] most important disputes, and would work
systematically against plaintiffs. The hypothetical case of the definitive study coming in two
years is obviously extreme. Both the likelihood of definitive resolution and the immediacy of
the prospect of such resolution will be much less threatening in real situations. In the context
of the Agent Orange Litigation there seems little prospect for a quick, definitive scientific
answer to whether (and to what degree) dioxin causes human cancer. This pessimism reflects
the great limitations under which such inquiry proceeds. Limitations on human
experimentation restrict scientists to examining the effects of toxic agents on humans who
have been exposed by mishap. Given the inevitable uncertainty about the circumstances of
individual exposures in real-world situations and the lack of homogeneity among humans,
there is a vast range for alternative hypotheses and hence little likelihood of firm conclusions.
Animal studies are useful in suggesting human risk factors, but scientists are extremely
cautious in extrapolating from animal studies to conclusions about causation in humans -- a
caution rooted in the obvious alternative hypothesis that humans are different. The upshot, in
the context of Agent Orange, is that the scientific picture of the role of dioxin in causing
human illness is likely to remain murky for a long time.
VI. THE DIFFICULTY OF DIFFERENTIATING CASES
A second problem that might seem to impede a verdict for Lilley is suggested by the blue bus
hypothetical. Dr. Carnow purported to be opining about Lilley as an individual, not about Lilley
as an undifferentiated constituent of an exposed class. The distinction is a matter of degree.
An expert's judgment about an individual will consider a variety of factors, each of which may
be thought of as defining classes, and all of which together may be thought of as defining the
class of persons presenting all such factors. At the same time, the essence of the difference
between a diagnostician and a statistician is that the diagnostician exercises intuitive
judgment about the relation of many factors, whereas the statistician is limited by his
methodology to considering but a few. The diagnostician will use detail specific to the
individual as a basis for opinion; the statistician is bedeviled by such detail.
Judge Weinstein challenged the adequacy of Dr. Carnow's knowledge about Lilley. Though
one can describe Dr. Carnow's opinion by reference to the standard of a treating physician, in
fact he did not treat Lilley. He saw Lilley's medical record only after Lilley had died, and the
medical record was rather thin since the doctors who did treat Lilley had not suspected dioxin
poisoning. End-of-service physicals contained no specific checks for symptoms of dioxin
exposure, and no provisions were made for reexamination in the future. Dr. Carnow had
learned of the dioxin exposure from Lilley's wife and brother-in-law. Although Judge Weinstein
considered Dr. Carnow's reliance on this information improper because it was hearsay (an
[*535] objection that seems unsound to me),Note 36 Judge Weinstein's real problem, I believe,
was that regardless of Lilley's specific medical history and legal position as an opt-out plaintiff,
his case was not an isolated tort, but rather was one small piece of a mass tort. Lilley was but
one among many soldiers exposed to dioxin. Neither Lilley's legal effort to opt out nor his
evidentiary effort to distinguish himself could persuade Judge Weinstein to differentiate him
from the class. When considering the problem of proving causation in this broader context,
perspective tends to change. Each of the many potential Agent Orange plaintiffs is unique in
physiology and background: each was exposed under different circumstances, and each
suffered different injury.Note 37 Yet the size of the class limits our ability to differentiate
individuals and to draw conclusions specific to each. Any class approach to the claims of a
population exposed to a toxic substance tends perforce to obscure the differences among
class members. For just such reasons, courts rarely certify mass toxic torts as class actions.
Notwithstanding the differences among the cases of each Agent Orange plaintiff, their number
and similarity make the task of trying each case separately by jury overwhelming.Note 38
[*536] In a mass tort we are forced to think of the population as a whole, which seems
naturally to entail addressing the question of causation in terms of the effect of dioxin
exposure on the population. Since that is just the kind of question statistical science
addresses, it is little wonder that Judge Weinstein thought the question should be answered
according to the methodology and standard of proof science would consider appropriate --
proof by epidemiological study. Because the class plaintiffs had no experiments or statistical
studies proving that dioxin caused cancer in the population of soldiers exposed to Agent
Orange, they had no case.
If that stark conclusion was right for the class, then Judge Weinstein was virtually compelled
to rule against Lilley. The class action and the individual actions were logically related.
Imagine what would have happened had Lilley won. In that event Judge Weinstein would have
determined that a rational jury could have concluded under the civil standard of proof that
Lilley's exposure to Agent Orange caused his injuries. But if a jury could rationally reach that
conclusion for Lilley, must it not also follow that a jury could rationally conclude that dioxin
caused cancer among the population of soldiers exposed in Vietnam? Were Judge Weinstein
to allow a medical diagnostician's opinion to satisfy Lilley's burden of proving causation, he
could not refuse to allow a similar expert opinion to satisfy the burden of proof for the class,
regardless of what epidemiology had to say.
The individual and class actions were also practically linked. Under the terms of the
settlement agreement, the defendants reserved the right to reject the settlement if a
"substantial" number of class members opted out.Note 39 Settlement of the mass of cases thus
depended on not too many plaintiffs opting out of the class. But a victory for Lilley would have
shown that opting out was the way to go. If opt-out plaintiffs could reach juries, they could
expect awards much larger than their participation in the mass settlement would provide. Their
cases would, therefore, have a correspondingly higher settlement value.
Thus there were pressures, both intellectual and practical, pushing Judge Weinstein toward a
population-based, statistical approach to the standard of proof -- an objective, probabilistic
conception of the standard of proof that plaintiffs could not meet. But the pressures did not
compel the approach. Judge Weinstein could have proceeded with the class just as he could
have proceeded with Lilley. He said he was prepared to try representative cases. Had he done
so under a standard of proof that allowed the plaintiffs' experts to testify and the cases to
reach juries, some plaintiffs would undoubtedly have lost; some may have won. The jury
verdicts would have begun to create a valuation landscape, and thus to provide a basis for
negotiating settlements.Note 40 A long road would have loomed ahead (which Judge Weinstein
[*537] avoided by his settlement strategy), but it is a road that would have preserved the jury
as a fundamental source of justice in resolving difficult disputes.
VII. BALANCING CONCERNS ABOUT THE ACCEPTABILITY OF VERDICTS?
Judge Weinstein's insistence on a statistical demonstration of probability has much to
recommend it in terms of acceptability. We should judge his approach not by the settlement
in the Agent Orange cases, but rather by what would have resulted had Judge Weinstein
dismissed the class action for failure of proof -- complete foreclosure of the plaintiffs' claims.
Wiping out the plaintiffs' claims on the basis of the burden of proof eliminates the risk that the
court will rest verdicts on a factual premise that is subsequently proven wrong. The force of
this consideration depends on one's assessment of the likelihood that science will soon prove
that dioxin does not harm humans. Wiping out the plaintiffs on the basis of the burden of proof
also eliminates the prospect of inconsistent jury findings about causation in the trials of
representative cases. The force of this consideration depends on one's assessment of the
system's capacity to rationalize inconsistent verdicts in terms of credibility of witnesses,
ability of lawyers, variations among juries, and similar considerations. Surely the more just
plaintiffs' recoveries seem, the less concern there will be about inconsistency. A verdict for a
plaintiff like Lilley would seem just, and thus acceptable, because the defendants acted
wrongfully in exposing him to dioxin, and it is believable that dioxin caused his injuries even if
it is not statistically provable.
Application of the traditional expert standard, which would have allowed Lilley's case to be
tried to a jury, raises serious concerns about the acceptability of resulting verdicts. Such
concerns, however, do not seem overwhelming, and should be balanced against the negative
consequences of Judge Weinstein's approach. Requiring statistical proof substantially
insulates companies from the consequences of negligently exposing persons to toxins.
Making proof by statistical study a requisite for plaintiffs in toxic tort litigation hamstrings the
ability of the judiciary to play a constructive role in future controversies. The next time a
situation like Agent Orange arises, the defendants will know they are not greatly at risk.
Instead of an incentive to care, companies are given an incentive to ignore even known risks.
Their major task will be to use what resources and political muscle they have to obstruct or
skew any epidemiological study that might prove statistically that their conduct caused injury.
Epidemiological studies are by no means a neutral tool equally available to both sides in mass
tort situations. They are huge undertakings that require the resources of large institutions like
corporations or governments, resources not within the means or ability of plaintiffs' lawyers.
Even when well conducted, their power to prove causation is often limited because of the
variety of alternative hypotheses and the amount of background.
[*538] Such studies are, moreover, prone to bias. Judge Weinstein was most impressed by
the study done by the United States Air Force, calling it "one of the most intensive
examinations to date of the effect of Agent Orange on exposed veterans."Note 4 The Air Force
study, known as the Ranch Hand Study after the code name of the Agent Orange spraying
program in Vietman, concluded with a cautious, meticulously scientific negative: "[T]here is
insufficient evidence to support a cause and effect relationship between herbicide exposure
and adverse health in the Ranch Hand group at this time."Note 42
The study suggested a 20-year follow-up. Judge Weinstein, by relying on the Air Force study
as a basis for throwing the veterans and their families out of court, ignored the problem of bias,
and thereby foreclosed an obvious question of weight and credibility normally left for the jury.
The study was conducted by one of the major wrongdoers, the Air Force,Note 43 which took
advantage of its men in the field to gain a public relations advantage against the electorate
back home. Epidemiological studies are by no means immune to bias: the more healthy
people included in a study population, the lower the statistical significance of injuries will
appear. How does a study determine who was exposed? The Air Force study included
everyone in the fixed-wing spraying program including pilots, navigators, and others not as
likely to have been exposed to Agent Orange as are those who were pouring Agent Orange
from drums to tanks. The Air Force study itself advertises its bias by evaluating the
significance of its conclusion. In highly unscientific fashion, the study makes a propaganda
statement (quoted by Judge Weinstein without any apparent recognition of the bias it
suggests): "In full context, the baseline study results should be viewed as reassuring to the
Ranch Handers and their families at this time."Note 41 Judge Weinstein backed up his reliance
on the Air Force study with a similar study done by the Australian Government, which had a
similar problem with its men in Vietnam.Note 45
VIII. CONCLUSION
A requirement that proof of causation be made exclusively according to the standards and
methodology of statistical science seems wrong to me both in the individual and the class
context. The requirement is tantamount to [*539] institutionalizing an objectively
determinable probability greater than .5 for cases in which proof of causation involves a
disputed supposition about the working of nature.
My object, though, is not to persuade you one way or the other about the result in Lilley. My
intent is to demonstrate the aptness of analyzing evidentiary problems and evidentiary
standards in terms of the judicial task of producing acceptable and stable verdicts, and further
to demonstrate that acceptability is not a simple matter of mathematical probability, but rather
a complex balancing that requires judgment about the objectives and mechanisms of the trial
process. Some may think that only statistical proof provides an acceptable basis for drawing a
legal conclusion about causation in cases of cancer and birth defect, while others would
accept conclusions based on less objective evidence. My point here is that acceptability is
the key.
