|
A. The Nature of Expert Testimony
Whatever the subject matter of their testimony, experts testify, expressly or impliedly, to
factual claims about the real world - that is, categorical propositions. Categorical propositions are
either true or false, 1 and which they are depends on the way the world is. This is what Professor
Allen was probably alluding to when he remarked that “testimony at trial must rest on
something...and that ‘something’ must be true”. 2 That “something” also is, or can be reduced
to,
“facts”. To avoid ambiguity, it is important to distinguish the different types of factual claims
experts can make because the mental processes implicated in forming a belief that the facts
asserted are true depend on the nature of the factual claim being asserted.
(a) Apprehended Facts: sense data, measurements, and derived facts.
The kind of facts lay witnesses testify about are perceptions, sense data, directly
observable events, the most basic of empirical facts. Ever since empirical facts were accepted as
the only source of knowledge about the real world, 3
only knowledge derived from sense data was
deemed to carry “the affidavit of truth”. 4
Some “facts” - my blood pressure or cholesterol level, the ambient temperature or
pressure - cannot be observed directly, only indirectly. Indirectly observable facts are linked to
1 ALFRED TARSKI, INTRODUCTION TO LOGIC
AND TO THE METHODOLOGY OF DEDUCTIVE SCIENCES (1941)
(“Every scientific theory is a system of sentences which are accepted as true... or, for short, simply
statements”);
S USANNE K. LANGER, INTRODUCTION
TO SYMBOLIC LOGIC (1953) p 71 (noting that propositions have “truth
values”, that is, they map unto a dichotomous variable that takes one of two values, “true”
or “false”).
2 Allen, supra note 28, at 6.
3 ALFRED JULES AYER, LANGUAGE, TRUTH
& LOGIC (1946).
4 SUSANNE K.
LANGER, PHILOSOPHY IN A
NEW KEY (1956)
p15.
sense data by some “translation system”, which usually consists in some sort of instrumentation. 5
They may be called “measurements” with the understanding that facts of this sort are measured
not only on an interval scale, but also on ordinal and binary scales. A chest X-ray showing
consolidation would be an indirectly observable fact or “measurement” in this sense.
Still other “facts” - for example, the inflation rate, the rate at which the concentration of a
hormone is increasing in a patient’s blood or whether a particular advertisement confuses
consumers - cannot be observed or measured, only derived. That is, they can be apprehended but
only by calculation or experimentation of some sort using direct or indirect observations. These
facts describe instantiations of an event or “state of affairs”. The claims are not evaluative or
epistemic but ontological in nature. They make a specific claim about the way the world actually
is. Thus, for example, a claim that an advertisement is confusing to consumers, is not an
evaluative statement about a particular advertisement, much less some general claim about how
consumers interpret advertisements. It is a factual claim that a particular advertisement confuses
consumers, defined as the proportion of the population exposed to the advertisement that will
misinterpret its intended meaning.
Depending on the context, derived facts may be referred to as calculations or estimations.
The inflation rate, for example, is estimated, whereas the rate of hCG increase in serum must be
calculated from serial measurements of the hormone. The important point, however is to
distinguish derived facts from facts that are inferred by induction, a subcategory of an entirely
different class of facts, to be discussed next, that are not apprehended in the three ways just
described but inferred from those that are. The distinction is more than semantic, as can be
illustrated by comparing two types of factual statements: the inflation rate today is 4% and the
5 Risinger, supra note 18, at
518-26.
five-year survival rate from stage III ovarian cancer is 35%. Although the estimation of both
involves extrapolation from sample to population, implicit in the latter is an inductive claim.
When one estimates the inflation rate, one is making a claim about the state of the world
as it is when the estimation is made. The estimation involves extrapolation from sample to
population, but the mental leap stops there. The claim is a claim about what the rate is at a given
time, it is not a general claim about inflation applicable to all inflation rates or even to rates that
obtain under a given set of circumstances. The estimate, like the apprehension of all types of facts,
is subject to measurement/perception error. And it may be invalid/biased if the method of
sampling is invalid/biased, just as a cholesterol value may be invalid/biased if the device measures
not only cholesterol but all steroids in blood or gives systematically high or low readings. But, if
an estimate or measurement is methodologically valid, the claim that it is what the result indicates
it to be is valid, and no other considerations affect the validity of the result. The inflation rate may
be affected by many factors, and a few months from now it may be slightly different, but when we
say that the inflation rate is 4% today, it is a true statement within the limits of experimental error,
if the method of estimation was valid.
The situation is rather different when we make a factual claim like, “the five-year survival
rate from stage III ovarian cancer following debulking surgery and combination chemotherapy is
35%”. This statement is not intended to be a factual claim about a specific population at specific
point in time. The claim reaches beyond the particular population of patients with stage III
ovarian cancer from which the sample was drawn to all patients with stage III ovarian cancer who
are alike in other respects to the trial subjects, that is, to all such patients in the future as well. It
is, in other words, a claim of about the state of things in general, not a specific factual claim about
a specific population.
These, then, are the three kinds of apprehended facts expert witnesses testify about: sense
data (perceptions, directly observable facts); indirectly observable facts (measurements on an
interval, ordinal or binary scale); derived facts. The “validity” of these facts will be influenced
by
the manner in which they were apprehended or “methodology”, where methodology is taken to
mean not only the means by which observations are made but also the design of experiments and
methods of analysis used. The validity of sense-data depends on the accuracy of perception; the
validity of indirect observations depends on the accuracy of the translation system that converts
sense data into the facts we seek to know; the validity of derived facts depends on the manner in
which we make our observations - the experimental design, the methods of analysis, as well as the
accuracy of the observations themselves. Most facts that expert witnesses testify about are not
directly observable sense data; 6 they are indirectly observable or derived facts.7
6 Experts would testify to sense
data if, for example, they examined the plaintiff or the scene of an accident
or fire, or if they were also a treating physician. Some purely experiential expertise is also based solely
on sense
data - such as an art critics opinion about the originality of a painting, or a chicken-sexer’s judgment
about the
gender of poultry.
7 Professor Langer’s unsurpassed
description of the indirect nature of scientific observation (written in
1951) applies equally to all non-sense data, and is worth quoting:
“Observation has become almost entirely indirect; and readings take
the place of genuine
witness. The sense-data on which the propositions of modern science rest are, for the most part, little
photographic spots and blurs, or inky curvy lines on paper. These data are empiric enough, but of course
they are not themselves the phenomena in question; the phenomena stand behind them as their supposed
causes. Instead of watching the process that interests us, that is to be verified...we really see only the
fluctuations of a tiny arrow, the trailing path of a stylus, or the appearance of a speck of light, and
calculate to the ‘facts’ of our science . What is directly observed is only a sign of the ‘physical fact’; it
requires interpretation to yield scientific propositions. Not simply seeing is believing, but seeing and
calculating, seeing and translating ...
The problem of observation is all but eclipsed by the problem of meaning. And
the triumph of
empiricism in science is jeopardized by the surprising truth that our sense-data are primarily
symbols ...Here, suddenly, it becomes
apparent that the age of science has begotten a new philosophical
issue, inestimably more profound than its original empiricism: for in all quietness, along purely rational
lines, mathematics has developed just as brilliantly and vitally as any experimental techniques, and, step
by step, has kept abreast of discovery and observation; and all at once, the edifice of human knowledge
stands before us not as a vast collection of sense reports, but as a structure of facts that are symbols and
laws that are their meaning ”.
(Italics in the original). See
Langer, supra note 102 at 20-21.
(b) Inferred Facts: “theories”, “principles”, “hypotheses”, “opinions”,
“conclusions”, “inferences”.
There is an entirely different category of facts from those just described - facts that are not
apprehended but inferred from those that are. Depending on the context in which they are used,
inferred facts, or simply “inferences”, are usually referred to, as “conclusions”, “opinions”,
“theories”, “principles” and “hypotheses”. 8 For example, factual claims that are tested by an
experiment are usually called hypotheses, and the inferences drawn from the results of the
experiment are usually called “conclusions”. Hypotheses and conclusions may, therefore, refer to
the same factual claims depending on the context.
Inferred facts are the product of two very different mental processes, called induction and
deduction. Colloquially, both are referred to as “reasoning”, but they are not interchangeable and
lead to conclusions that have quite different properties.
Inferences are the conclusions of “arguments”. Argument is a term of art that refers to an
assertion that certain statements, called conclusions, follow from other statements, called
premises. To say that a conclusion follows from the premises of an argument is to say that if the
conclusion is true it is because the
premises are true. The difference between deductive and
inductive arguments lies in the relationship between the premises and conclusions of these
arguments, that is, in the sense in which the conclusions are said to “follow” from the premises
of
the argument.
In the case of deductive arguments, there are rules that make such arguments valid. These
are the rules of formal logic. What it means to say that a deductive argument is valid
is to say that
these formal rules are satisfied. And if the formal rules are satisfied, the conclusion of the
8 Daubert, 509 U.S. at 592 (... “ideas inferred from... facts [that
are known]...”)
argument must be true if the premises
are true. This may or may not be a big “if”, but if the
premises are true and the form of the argument is valid, the conclusions are compelled
by the
premises. This is so because the conclusion of a valid deductive argument are entailed
by its
premises , a fact that has important (and surprising) corollaries.
If the conclusion of a deductively valid argument is entailed by its premises, it cannot
contain more information about the world than the premises. And if all the information about the
real world contained in the conclusion is already contained in its premises, we cannot discover
new information about the world deductively - i.e. through formal logic - we can only discover
new insights. This is turn means that we cannot deduce the truth of the premises
of an argument
logically unless the premises themselves are the conclusions of logical arguments. A chain of
logical arguments may be constructed where the conclusions of one is the premise of a subsequent
argument, but eventually primary premises are reached whose truth cannot be determined
deductively (logically).
The conclusion of an inductive argument contains claims about unobserved facts that are
inferred from observed ones. These claims assert facts not contained in the premises (the observed
facts), and so the conclusion cannot be entailed by the premises, nor is the conclusion compelled
by the premises.
The conclusions of inductive arguments assert generalizations that are said to follow from
the premises because the observed facts are taken to be instantiations of the generalizations
asserted in the conclusions. Inductive arguments provide new information about the real world,
but because the conclusions of these arguments are not entailed by the premises the conclusion is
never compelled. There can, therefore,
never be a guarantee that the conclusion of an inductive
argument is correct, there can only be varying degrees of probability that the conclusion is true. 9
Uncertainty is the price of new information. 10
Whether an inductive inference is “true”, or, more properly, how true it is, is a
functionsolely of the instances of the particular observations on which it is based. For a
conclusion, c, to follow rationally from the evidence, e, on which it is based, it must be true that,
given the evidence, c is more likely to be true than not true: symbolically, that p(c *e) > p(~c*e)
(where ~c means “not-c”). This is obvious, for if not-c were more likely than c, given the
evidence, then we should conclude not-c instead of c. That p(c *e)
> p(~c*e) is the minimal
requirement for a warranted belief that c is true, given evidence, e.
The notion that logic can only provide insights based on facts that are already known, but
not new information about the world, 11
seems counterintuitive, and at first blush suggests that
looking at the logic of an experts “reasoning” cannot be a very helpful way to determine whether
or not his conclusions are correct. In fact, courts have had the least difficulty barring expert
testimony when the expert’s conclusions do no follow from his or her premises. The reason is that
when the conclusion does not follow from the premises, it means, formally, that the argument is
deductively invalid. Therefore, it is immaterial whether the premises are true or not because the
9 See e.g. ROBERT
AUDI, ED.,
THE CAMBRIDGE DICTIONARY
OF PHILOSOPHY (1999);
KEVIN T. MASLIN,
A N INTRODUCTION
TO THE PHILOSOPHY OF MIND (2001).
10 Where two events are causally
linked by strict physical laws, however, the occurrence of one will
predict the other with “certainty”, not merely make it occurrence more likely than not. These are
the so called laws
of nature of which the court can take judicial notice.
11 The new insights discovered
logically are called by philosophers tautologies. Logical tautologies
do not appear to us as tautologies - i.e. as self-evident - only because of our limited intelligence. For example,
it is
not self-evident to us that 81x69= 5,589. We have to carry out a calculation to see that this is so, but the
statement,
81x69= 5,589, is nonetheless a tautology for that because 5,589 provide us with no information about the world
that 81x69 does not provide. The same is true for more complicated logical and mathematical relationships. See
Ayer, supra note 101, at Chapter IV.
conclusion cannot be true because the
premises are true. Thus, when an expert’s arguments lead
to contradictions or conclusions that do not follow from the premises, the court’s task is straight
forward.
Invalid deductions usually consist in one of two fallacies: affirming the consequent and
denying the antecedent. 12 An example of the fallacy of affirming the consequent is the suggestion
that courts should not categorically reject animal studies as evidence of causation in toxic tort
cases because all of the 40 or so chemicals recognized to cause cancer in humans also cause
cancer in animals. 13 The unstated form of this argument is the conditional, “If a chemical is a
human carcinogen, it will be an animal carcinogen; a substance is an animal carcinogen, therefore,
it is a human carcinogen” (and so courts should not reject animal studies). Not so.
Symbolically, the fallacy can be written, “If A, then B; B; therefore, A”. The reason for the
fallacy is that the conditional does not state that A is the only cause of B, but the conclusion
assumes that it is. There are many substances that are carcinogens in animals that are not
carcinogens in humans. 14 Therefore, nothing can be concluded logically about
the probability that
a compound is carcinogenic in humans from the fact that it is carcinogenic in animals and that all
human carcinogens are animal carcinogens.
Denying the antecedent is a fallacy encountered in trying to draw conclusions about
causation in toxic tort cases using the “method” of differential diagnosis. 15 The idea here is that if
12 Supra note 107.
13 Beeches-Monas, supra note 20, at 1065-66.
14 An example that caused so much
uncertainty among gynecologist is Provera, a compound that
causes mammary cancers in Beagle dogs, and is one of the mainstays of therapy of dysfunctional bleeding.
15 Michael B. Kent, Daubert, Doctors and Differential Diagnosis: treating
Medical Causation
Testimony as Evidence , 66 DEF.
COUNS. J. 525 (1999).
a drug is known to be associated with an illness, and other known causes of an illness can be
excluded, it follows that the drug in question must have caused the illness. Stated formally, the
argument is as follows: If A or B or C are present, D did not cause the illness; A, B or C were not
present, therefore, D caused the illness.
Symbolically, if P stands for the antecedent proposition and Q for the consequent, the
argument is of the form: If P, then Q; not P, therefore, not Q. The fallacy lies in the assumption
that A, B, C, and D are the only causes of the illness in question – i.e. that all the causes are
known, which is rarely if ever the case. The causes of most sporadic illnesses are not known.
A proposition containing a factual assertion may, therefore, be true or false in three
senses: (1) the fact asserted may not be what the expert claims it to be, i.e. the expert simply “got
his facts wrong”; (2) the facts inferred may not follow deductively or inductively from the
premises on which they are purportedly based; or (3) the methodology used to apprehend the fact
- be it the instrument used for measurement or the experimental design by which the facts were
derived - was sufficiently deficient to render the “fact” no more likely to be true than untrue.
To
determine whether such claims are true or false requires a historical inquiry in the first case, a
logical inquiry in the second case, and a scientific inquiry in the third case. Courts are eminently
qualified to conduct at least the first two of these inquiries.
|
