Biology Reference
In-Depth Information
Cole, 2009
), that of matching perhaps slightly less so, we therefore proceed based on the
simple discussion above.
In comparing two items of impression evidence based on shape, the first step is to
determine the level of variation present in repeated measurements of the same specimen.
A series of shape (or size and shape) measurements is made of a set of impressions pro-
duced by a single specimen. The variability within this set of repeated measures may be
characterized by the variance within the collection or by looking at the distribution of all
pair-wise distances between specimens in the repeated measurements group to develop an
estimate of a confidence interval of distances within repeated measurements. If the dis-
tance between two shapes (measured on different impressions) lies within this repeated
measures range of values, then there would appear to be little difference in shape, and
hence no reason to claim that the impressions could not be have been produced by a com-
mon source. Detection of the “match” is thus based on a failure to reject a null hypothesis
that the observed difference between the two arose from random variation in the measure-
ment process. Unfortunately, this leaves us in the rather precarious position of accepting a
null hypothesis, of simply stating that one cannot tell the objects apart.
Certainly we need to be cautious when taking this approach. For example, it may be
that virtually all the difference between two shapes is localized to one landmark, inducing
a Pinocchio effect. Because Procrustes methods tend to map variance across the whole con-
figuration, a substantial change at one location might be mapped onto smaller changes at
many landmarks. The result could be a distance between the two shapes comparable to
one observed in repeated measures even though a large excursion at just one landmark is
highly unlikely in repeated measurements. In such situations, Procrustes plots could be
examined to check that the distribution of relative shape changes across landmarks
produced by repeated measurements is consistent with the observed Pinocchio effect
difference.
Using Procrustes (or Procustes SP) distances in the match criteria, albeit subject to con-
cerns about the Pinocchio effect, allows for automated searches for matches in databases
of specimens. It is straightforward to compute the Procrustes distance between a given
specimen and all specimens in a database, and then to determine the number of matches
found, given the match criteria.
MATCHING SHAPES IN THE HUMAN DENTITION
Use of bitemark analysis in criminal cases arose out of the highly successful practice of
using dental records to identify the remains of individuals where extensive damage or
decomposition has made it difficult to establish identity. Bitemark analysis is used to
investigate particularly violent events, typically sexual assaults and child abuse, making
this application a highly sensitive topic, not only because false identifications could
imprison an innocent person, but also because a violent offender may be left at large. The
analysis of bitemarks yields a range of forensic information in these disturbing crimes, but
there are real concerns about its use as individuating evidence, i.e., as evidence uniquely
linking a suspect to a bitemark. The National Academy of Sciences report (2009) noted
that the unique association of a suspect with a bitemark requires that: (1) the human biting
