8127/frame/fm

 
c.
The Analyst and the Evidence — Articulating Hypotheses
Regardless of which statistical framework one chooses, some person must
articulate the hypothesis or hypotheses to be tested. The statistics are only a
tool to evaluate the hypotheses. If stupid or irrelevant hypotheses are pro-
posed, the statistics will evaluate those just as easily as they will intelligent
and useful hypotheses. If the correct hypothesis is never proffered, no statis-
tical or logical framework can rescue the examination from defeat.
To formulate relevant hypotheses, the analyst must have as many case
details as will assist in this effort, and he may consult any number and variety
of individuals and sources to achieve this goal. This is one of the primary
reasons that a criminalist should respond to crime scenes: to gain firsthand
knowledge of the problem. Provisional hypotheses can be made on the spot,
and relevant evidence collected. In addition, reading police reports, discus-
sions with detectives and attorneys, and knowledge of other physical evidence
results all contribute to a thoughtful framing of hypotheses. Creativity is also
useful in imagining what might have happened. Most important, each
hypothesis must be falsifiable, and some means of evaluating a failure to
falsify must exist. All hypotheses should be directed to answering some legal
fact in dispute, whether it be the contact of the suspect with the scene, contact
between victim and suspect, or even contact of the victim with a crime scene.
8127/frame/ch07 Page 172 Friday, July 21, 2000 11:46 AM

---

Association and Reconstruction — Inference of Contact
173
 
d.
Absence of Evidence
We include a discussion of absence of evidence in this chapter because this
question usually applies at the level of 
 
association
— did contact occur
between the putative target and a suspected source? For example, a rape
victim describes a green wool jacket that the perpetrator was wearing during
the incident. A jacket of that description is found in the suspect’s car; however,
no green wool fibers were recovered from the victim, her clothing, or the
immediate crime scene. Does the failure to find “expected” evidence support
a hypothesis that the jacket (
 
source
) and the victim (
 
target
) were not in
contact? Strictly speaking, the failure to find evidence does not necessarily
lead to a conclusion that no contact occurred. This is a difficult problem in
hypothesis testing. The null hypothesis is, “the source and target were in
contact.” The scientific question is, can a negative result falsify a hypothesis?
This depends heavily on our knowledge about certain properties of the item
in question, such as transferability, persistence, and detectability. As we dis-
cussed in the previous chapter, what knowledge we have of these attributes
may be limited and neither accurate nor precise.
Nevertheless, the bottom-line question remains whether we would ever
accept an absence of evidence as evidence of absence. Many analysts do not
believe a lack of evidence to be strong evidence that no contact occurred.
They have adopted a more specific application of the general maxim,
“Absence of evidence is not evidence of absence.” How do we evaluate evi-
dence when there isn’t any? Scientists are accustomed to analyzing material
things; they are not accustomed to evaluating the absence of something. One
way of looking at this problem is that although nothing may not 
 
be
some-
thing, it may still 
 
mean
something. One way of thinking about a problem
such as this is to take it to a logical extreme.
Suppose a man attacks a woman. She is able to describe him as wearing a
white shirt and jeans. She sprays him with ortho-chlorobenzilidinemalano-
nitrile (tear gas) with an orange dye mixed in. A man wearing a white shirt
and jeans is picked up running from the scene. An examination of his shirt
shows no traces of dye.*
Under this scenario, the failure to find the orange dye (in other words, a
finding of nothing) provides substantial evidence supporting the hypothesis
that the arrestee is not the attacker. We have a very high 
 
a priori 
expectation
that orange dye will be present on the shirt of the attacker.
* This example was offered by Dr. John Buckleton during an exchange on an Internet e-mail
list.
8127/frame/ch07 Page 173 Friday, July 21, 2000 11:46 AM

---

174
Principles and Practice of Criminalistics
A Bayesian framework allows us to compare the likelihood of contact vs.
no contact rather than depending on the unsatisfying exercise of trying to
nullify a hypothesis using a negative result. In Bayesian terms, the evidence
(lack of orange dye on the suspect) is so much more likely if he were not the
person sprayed, than if he were; this evaluation overcomes almost any prior
odds that we have identified the correct suspect. The resulting posterior odds
tell us we have the wrong guy. Note also, that the assumptions necessary to
associate a non-orange shirt with an innocent suspect under this scenario
are relatively few. We must assume he did not have time to change the shirt,
as appears likely from the scenario, we must assume that the woman’s aim
at point-blank range was good enough to hit the attacker, and we must
assume that she is in fact telling the truth. That we have a reasonable expec-
tation in this circumstance about what an absence of evidence reveals suggests
that we might also have expectations in more ambiguous cases. However, the
more assumptions we must make about why no evidence exists, the weaker
will be any inferences that explain the evidence.
An interesting example using absence of evidence to infer one kind of
contact rather than another can be found in Sidebar 5. In this example, the
behavioral analyst (profiler) suggests that, based on a lack of transfer of
Marilyn Sheppard’s blood to Sam Sheppard, and subsequently to items he
may have touched, he must have been her killer. It is interesting to note the
unarticulated assumptions in this analysis. They include:
1. That he successfully washed all the blood away after killing her,
2. That he couldn’t have washed all the blood away after touching her
to take her pulse,
3. That all the blood was hers,
4. That the blood on the watch is hers, and
5. That the bloodstain on his pants knee is irrelevant.
The final 
 
coup de grace
that makes the entire exercise absolutely worthless is
the 
 
assumption
that Sam Sheppard was the killer. We encourage readers to
examine the entire analysis (McCrary, 1999) and reach their own conclusions.
The temptation in evaluating negative results is to dismiss the failure to
find evidence as insignificant, and therefore of no consequence to the case.
The two examples provided above represent opposite ends of the spectrum
when considering the implication of an absence of evidence. As for all asso-
ciative evidence, the relevance and strength of the inference rests entirely on
the facts and assumptions from which it is formed. The fewer facts we have
and the more assumptions we are then forced to make, the weaker the
inference. Situations involving an absence of evidence frequently offer fewer
facts and necessitate more assumptions. As long as this is explicated, there is
nothing inherently wrong about offering a conclusion.
8127/frame/ch07 Page 174 Friday, July 21, 2000 11:46 AM

---

Association and Reconstruction — Inference of Contact
175
An inference stemming from absence of evidence necessarily relies
heavily on the experience of the examiner. Some practitioners are uncom-
fortable with this. They also believe that the variables involved cannot be
adequately enumerated, controlled, and quantified. We suggest that an expe-
rienced examiner might give an opinion that he would have expected to
detect traces of the source on the target, and that their absence, therefore,
provides useful information. Such an opinion should be based on as much
supporting data as possible, and the assumptions should be explicated. Obvi-
ously, no consensus exists in the field regarding this issue. The challenge to
the forensic community is to address this question, both from a research and
intellectual perspective, and to continue the discussion.