Report of the Research Project (36 EC)
Master of Science in Forensic Science, University of Amsterdam July – December 2019
Optimizing trace examinations using
Bayesian Decision Networks
by
Katharina Draxel Student-ID: 11852046
Conducted at the Netherlands Forensic Institute
Supervisor: Dr. Yvonne van de Wal, Netherlands Forensic Institute
Examiner: Dr. Maarten Blom, University of Amsterdam & Netherlands Forensic Institute Submitted on the 10th of December 2019
Abstract
Adhesive tape is a common piece of evidence that is used for a wide array of crimes. Its trace examination is a challenging task due to the variety of traces and the complex trace dynamics associated with the tape. Additionally, as the assessment on activity level gains more importance, a different examination would be necessary to provide information on an activity. This trace examination of adhesive tapes involves a multitude of decisions. These decisions can highly influence the outcome of the examination, and optimal decisions have to be made to ensure a successful examination.
Interviews with fingermarks and DNA trace examiners and a case file study were conducted to gain insight into the trace examination of adhesive tapes and the involved decisions. The obtained information could be used to model a Bayesian Decision Network (BDN). Such a BDN uses Bayesian decision theory to quantify the desirability of outcomes and calculate the optimal decision based on data.
In this study, a BDN was constructed for the decision of the sampling strategy for contact traces of DNA trace examiners. It visualizes the influences for the decision and provides the optimal sampling strategy under different case circumstances. This is based on success rates and the utilities assigned to the outcomes. Additionally, the network can account for the differences in sampling for source and activity level questions. Therefore, the analysis of a decision using a BDN facilitates simplifying complex decisions and making rational decisions based on data. For that reason, it can be used to optimize protocols for the examination of forensic evidence.
Keywords: Decision-making, Bayesian decision theory, Bayesian decision network, sampling strategy, DNA contact traces, adhesive tapes
3
Introduction
During the entire examination of forensic evidence, from the trace recovery to the evaluation, a multitude of decisions are made. These decisions highly influence the outcome of the analysis, and it is, therefore, crucial to make the optimal decisions. A proposed way of making these decisions is based on rational decision theory: the decisions should be rational, and the best possible option should be chosen by calculating the decisions’ advantages and disadvantages[1]. This decision theory can be combined with Bayesian statistics to produce a framework of decision-making under uncertainty[2]. Uncertainty is an inevitable part of forensic science as the underlying truth is not known. Probabilities can be used to measure these uncertainties. Under this resulting Bayesian decision theory, the desirability or utility of the consequences of decisions can be determined. The goal is to choose the decision with the maximum expected utility[3]. To calculate the maximum expected utility probabilities of the occurrence of the uncertain events/states of nature in question, as well as the utilities assigned to the possible outcomes/consequences, are needed. A graphical way to depict these decision problems is a Bayesian Decision Network (BDN)[4]. These BDNs can support the scientist in making decisions by breaking down complex decision problems and calculating the expected utility for each decision.
In this study, the aim was to construct such a BDN for the decisions made during the trace recovery from adhesive tapes such as duct tapes and packaging tapes. Adhesive tape is a common piece of evidence that involves a multidisciplinary examination. Traces that can be found on tape are amongst others DNA, fingermarks, hair, micro-traces (namely fibers and glass), and explosive residues[5]. Additionally, physical end matching of the tape can be done to reveal the order in which several pieces of tape came from the roll. This variety of traces leads to many disciplines working together that have to take the subsequent recovery of other traces into consideration while making decisions. The decision-making is additionally complicated by the complex trace dynamics associated with tape. For example, due to its stickiness, an additional collection of unrelated traces after the crime might occur, and multiple layers of tape can lead to the relocation of traces within the tape[5]. The most common disciplines working together during the examination of tape are human biological traces and fingermarks. Their procedures can highly influence each other as some visualization techniques for fingermarks will not allow for subsequent sampling of DNA (e.g., crystal violet as it includes washing steps[6]). Additionally, DNA trace examiners might adjust their sampling area based on found fingermarks (e.g., presence of fingermarks indicates contact with a specific area and therefore sampling that area could have a higher chance of DNA recovery). Furthermore, the trace examination of a piece of evidence is one of the most crucial steps in the investigation, as subsequent analysis steps are dependent on its success. It is desired to conduct the examination as successfully as possible. Therefore, it is important to optimize the existing protocols for trace recovery. At this time, protocols at the Netherlands Forensic Institute are optimized by several methods. For the validation of new sampling techniques, the existing literature is reviewed, and in-house studies are performed. Additionally, case file studies on the success rates of recovering traces from certain areas on items and of the applied sampling techniques are conducted. The goal of this project, however, is to present a new approach for optimizing the trace examination of adhesive tapes by using Bayesian decision theory. This facilitates making better-informed and rational decisions.
Moreover, current protocols are aimed at optimizing the trace recovery regarding questions on source level. This level for evaluating forensic evidence concerns the origin of a trace[7]. However, these questions shift increasingly to the activity level as the court does not doubt anymore whether a trace stems from a person. Instead, the court is more interested in how a trace got to the scene of a crime[8]. Regarding adhesive tapes, this shift is caused by issues like the handling of the tape role prior to a crime. In such cases, where the activity that leads to the deposition of a trace is disputed, the exact position on the tape a trace was recovered from becomes more crucial. The different activities that led to the deposition of a person’s DNA at these specific locations need to be considered, and examination protocols should be adjusted accordingly. Therefore, this project additionally aimed at incorporating activity level questions in the BDN to optimize the examination under these types of questions.
5
Material and Methods
Information on the examination protocol and decisions
An overview of the entire examination protocol on adhesive tapes had to be established. This was done by reading the existing document on the protocol for the multidisciplinary examination of tape at the Netherlands Forensic Institute, joining the laboratory for trace examinations, reading past examinations in case files and consulting scientists from all disciplines that are involved in the examination of tape. These tasks gave insight into the set-up of the protocol as well as the different decisions that must be made during the examination. Furthermore, interviews with five fingermarks and eight DNA trace examiners were conducted. For the interviews, three real cases on adhesive tapes were picked. One case out of the three through a case file study identified categories were chosen: 1) a victim case, which is a case involving a victim that was tied up with the tape, 2) an object case, which is a case where tape was wrapped around an object, and 3) a post-explosion case where the tape was subjected to an explosion. Additionally, only recent cases involving packaging tape or duct tape were chosen. Ideally, at least in one of the cases, fingermarks were found on the back side and the sticky side of the tape, as this opens up more options for decisions. The trace examiners were asked to provide information on how they would conduct the examination in these cases and why they would take these actions. Specific questions about interesting matters encountered during the information collection on the protocol were asked subsequently. These interviews provided additional information on why certain decisions are made. This information was used to identify influences for decisions and the utility of outcomes that have to be represented in the Bayesian Decision Network (BDN).
During the interviews with trace examiners, several situations were identified where consultation of reporting scientists for biological traces would occur. Therefore, additional interviews with three reporting scientists were conducted. For the identified situations, the reporting scientists were asked to provide information on how they would decide to proceed with the examination and why they would take these actions. The gained information could additionally be used to model influences for decisions and their utilities in the BDN.
Constructing a Bayesian Decision Network
To model the BDN, the software HUGIN Researcher was used[9]. A BDN consists of three different types of nodes: the chance nodes representing the states of nature/events in question, the decision nodes representing the decisions that can be made at a certain point in time, and the utility nodes representing the utilities for the different outcomes[4].
The states of the decision node are based on the information gathered through the interviews with trace examiners. The values within the decision node are so-called initial policies, and values of 1 and 0 can be assigned to illustrate which decisions are possible under certain events. The decision node’s parent nodes have to be known at the time the decision is made[4]. Parent nodes are nodes that precede another node and point towards that node[9].
The other two types of nodes require data, namely the probabilities of the occurrence of events for the chance nodes and the utilities of the consequences for the utility node. Utilities are self-assigned weights to the consequences based on the preferences of the decision-maker[2,3]. They represent the value a certain outcome has to the decision-maker. One way to assign these weights is on a scale from 0 to 1, with 0 being the least desired consequence and 1 the most desired consequence[4]. It is, however, possible to choose any scale using much higher numbers or even assigning negative weights.
In this study, chance nodes were differentiated into three categories. For the first category, the probabilities of occurrence for the events consist of success rates, which were collected during
a case file study. The second category comprises chance nodes that have to be instantiated based on case information. Therefore, their probability tables do not have to be filled out. If their states are unknown due to a lack of case information, the case will be treated as if no specific information was given on the matter (e.g., if it is unknown how many witnesses and victims were involved, the case will be treated as if no additional witnesses and victims handled the tape). The third category includes chance nodes that summarize information from their parent nodes. Their conditional probability tables have to be filled in. Based on the different states of their parent nodes, only certain states are possible (e.g., if a victim was wounded and blood was observed on the tape, victim blood will be present on the tape).
The dependencies between all the nodes are based on the information collected through the interviews with the trace examiners and reporting scientists. After the collection of data through the case file study, the dependencies for the best DNA profiling result were verified using Chi-squared tests.
Case file study for the collection of success rates
For the case file study, only duct tapes and packaging tapes were considered. Any other type of tape like isolation tape, double-sided tape, and also entire tape rolls are examined differently and were therefore excluded to reduce the complexity of the network. Furthermore, only cases after the introduction of the new and more sensitive STR analysis kit were included, as this highly influences the amount of DNA that is detectable and, therefore, still able to yield a profile. Of 118 cases in this time frame, two-thirds included duct tapes and packaging tapes. This way data on 516 samples from 147 items (investigated tapes) in 80 cases were collected. For the results of the DNA profiling, the interpretation of the reporting scientist working on that case was adopted. No interpretation of the profiles was conducted. All the collected data were evaluated to yield probabilities in the form of success rates. In instances where no probabilities could be derived from case files, probabilities can be elicited[10].
To gather data on the sampling strategy for contact traces, the different locations for sampling adhesive tapes had to be considered. Fig. 1 shows an illustration of how the tape can be divided into different areas for sampling contact traces.
An original end denotes an end that was created by the perpetrator and could, therefore, give information on the person who initially tore the tape. Ends that were introduced later (e.g., through cutting the tape by ambulance staff while untying a victim) are non-original ends and not of interest. Each strip of tape initially must have had two original ends. Therefore, differentiation into the two original ends and the rest of the back side is possible. In general, the back side of the tape denotes the smooth side opposite the sticky side containing the adhesive. It is the side that is sampled for contact traces.
Figure 1: Possible area differentiation of a strip of tape for sampling contact traces. The tape can be differentiated into the areas close to the original ends of the tape (approximately 2-3 cm), the rest of the back side, or the entire back side of the tape without differentiating original ends and the rest of the back side.
Rest of back side
1
storiginal end
2
ndoriginal end
7
Results
Examination protocol for adhesive tapes
A flowchart of the entire examination of adhesive tapes was generated (see the supplementary materials Appendix A). This was based on information collected from the existing protocol for the multidisciplinary examination of tape, joining the laboratory, reading past examinations in case files, consulting scientists from different disciplines, and interviewing trace examiners and reporting scientists. The focus was put on the steps conducted by fingermarks and DNA trace examiners. For further information on the examination process and the decisions involved, see the supplementary materials Appendix A.
Selection of decisions for constructing a Bayesian Decision Network
The first analysis of the examination protocol resulted in identifying a multitude of decisions that are made during the examination of tape for fingermarks and DNA. Two of these decisions were chosen to model a Bayesian Decision Network (BDN). The two chosen decisions concern the sampling strategy of DNA trace examiners for sampling contact traces. This includes the decisions where to sample (original ends/rest of the back side of the tape/the entire back side) and how to sample (separately/together over several strips of tape or regarding the original ends of one strip of tape). These two decisions were combined into one decision as they are highly intertwined. This decision about the sampling strategy for contact traces represents one of the most important and complex decisions in the protocol.
Constructing the Bayesian Decision Network
For the chosen decision of the sampling strategy for contact traces, a Bayesian Decision Network was constructed. The modeled BDN can be seen in Fig. 2. The network was constructed using information from the interviews with trace examiners and reporting scientists as well as the case file study. For further results from the interviews and the case file study, see the supplementary materials Appendix A, B & C.
Figure 2: Bayesian decision network for the decision on the sampling strategy for contact traces. Grey nodes depict chance nodes that are based on the case information and have to be instantiated. White nodes depict chance nodes that summarize information. Yellow nodes are chance nodes that contain success rates from case files. The red node depicts the decision node and the green node the utility node with the utilities for each outcome.
The decision
Contact traces sampling strategy node
At the center of the network is the decision node labeled Contact traces sampling strategy (shown in red in Fig. 2). This node contains the different sampling strategies that were identified during the interviews and case file study. All possible states of this decision node are listed in Table 1. These states consist of all possible combinations of only sampling original ends together or separately over all available strips or per strip, sampling the entire back side of the tape over all strips or per strip and the same options for sampling the original ends and the rest of the back side apart from each other. There is no state of no sampling, as it is assumed that an investigation should be done as the item was deemed useful by the police and sent in for investigation.
Table 1: Possible decisions for a sampling strategy for contact traces. The table shows all the possible states (d1-10) for the decision node “Contact traces sampling strategy”.
d1: Original ends separately over all strips d2: Original ends together over all strips d3: Original ends together per strip d4: Entire back side over all strips d5: Entire back side per strip
d6: Original ends separately over all strips & rest back side together over all strips d7: Original ends separately over all strips & rest back side per strip
d8: Original ends together over all strips & rest back side over all strips d9: Original ends together per strip & rest back side over all strips d10: Original ends together per strip & rest back side per strip
9 Influences on the decision
Following the identification of the decisions, their influences have to established. These influences are based on information from the interviews with trace examiners and reporting scientists. In total, four direct influences on the decision were identified. Each of these four influences is determined by certain other events. In this section, these influences and events are described in more detail and the described nodes are depicted in Fig. 2.
Amount of target and non-target DNA expected? nodes
One major factor that influences the decision is the amount of non-target and target DNA that can be expected on the tape. Target DNA describes the DNA left by a person of interest, while non-target DNA is not of interest for the evaluation. This factor is represented by the nodes
Amount of non-target DNA expected? and Amount of target DNA expected?. Both nodes have
the possible states of ‘none’, ‘low’, ‘medium’ and ‘high’. These states were kept more abstract as trace examiners do not think in actual concentrations during the investigation. Nevertheless, for collecting data on these probabilities, ranges of concentrations have to be chosen based on the data. However, data is not available for every combination of events that influences the amount of DNA. For this reason, success rates for this node are not yet included in the model. Probabilities can be elicited by experts to fill in these missing success rates.
Masking expected? and Sufficient amount of target DNA expected? node
There are two reasons why the amount of DNA expected on the tape influences the decision: possible masking of a low amount of target DNA by a high amount of non-target DNA and having, in general, a sufficient amount of target DNA. These parameters are represented in the nodes Masking expected? and Sufficient amount of target DNA expected?.
If masking is expected, more locations on the tape would be sampled separately to avoid collecting an excessive amount of non-target DNA in a sample. This could otherwise lead to an overpowering of target DNA by non-target DNA in a PCR.
If it is not expected to have a sufficient amount of target DNA on the tape, more locations on the tape would be sampled together in order to collect as much DNA as possible. This could give a higher chance of generating a DNA profile.
It should be noted that there are events that outweigh the effects of other events on the decision. For example, if masking is expected but also an insufficient amount of target DNA, more locations would still be sampled separately, as avoiding masking remains a priority. This outweighing holds for several other events and will be discussed again at a later point in this paper.
Victim bound with tape? and Type of case node
As already mentioned, several other events influence the amount of DNA that can be expected on the tape. The first event that influences the amount of non-target DNA on the tape is whether a victim was tied up with the tape or not. This is represented in the network with the node
Victim bound with tape? and can have the states ‘Yes’ and ‘No’. If a victim was tied up with the
tape, a high amount of victim DNA is expected to be present on the tape. This is due to the proximity and the intense contact of the victim with the tape.
In return, this is influenced by the type of case and is represented with the Type of case node. Its possible three states are ‘Victim case’, ‘Object case’ and ‘Post-explosion case’. These are the three main types of cases for which tape is used and were identified through case file studies. Per the definition employed in this study, only a victim case will result in a victim being bound with the tape.
Number of W or non-bound V that handled the tape? node
The next event that influences the amount of non-target DNA is the number of other non-target contributors like witnesses and non-bound victims that handled the tape. It is illustrated in the
Number of W or non-bound V that handled the tape? node and contains the three states ‘0’,
‘1’ and ‘2’. These are the possible numbers of additional non-target contributors that were identified during the case file study. Their presence influences the amount of non-target DNA as the more people handled the tape, the more DNA might be expected on the tape.
Victim blood present?, Blood observed? and Victim wounded? node
Another event that influences the amount of non-target DNA is whether victim blood is present or not. It is represented by the Victim blood present? node and contains the states ‘Yes’ and ‘No’. The chance of victim blood being present depends on whether blood was observed or not and whether it is known that the victim was wounded or not. This is illustrated with the
Blood observed? and Victim wounded? nodes respectively. Both nodes have the states ‘Yes’
and ‘No’. For example, if it is known that a victim was wounded and blood was observed on the tape, there is a very high chance that victim blood is present. When sampling for contact traces, blood will always be avoided to prevent contamination with a high amount of non-target DNA. However, this is a difficult task, and samples still often test positive for blood. Therefore, they could be contaminated with a high amount of non-target DNA from the blood. Blood that is suspected to stem from the perpetrator will always be sampled separately from contact traces.
Post-incident persistence node
A parameter that influences the amount of target and non-target DNA on the tape is the persistence of the DNA after its deposition. This is portrayed in the Post-incident persistence node and has the states ‘Low’ and ‘Sufficient’. The amount of DNA expected on the tape can be highly reduced if the tape was exposed to an explosion, other forms of extreme heat or water. In such cases, the persistence of DNA would be low, which would result in a low expected amount of target and non-target DNA. If no such extreme circumstances are known, it is assumed that the persistence is sufficient and has no major influence on how much DNA could be expected on the tape.
Gloves worn by perpetrator? node
Another event influencing the amount of target DNA expected on the tape is whether it is known that the perpetrator wore gloves or not. It is represented in the Gloves worn by perpetrator? node and has the states ‘Yes’ and ‘No’. A lower amount of target DNA will be expected if it is known that the perpetrator wore gloves.
Expected number perpetrators node
What might stand out is that for the amount of non-target DNA a number of contributors node is present, while this is not the case for the amount of target DNA. This is because the expected number of perpetrators has a direct influence on the sampling strategy. As soon as it is expected that more than one perpetrator is involved in the case, more locations will be sampled separately in order to avoid complex mixtures. Even if an insufficient amount of target DNA is expected, but several perpetrators are involved, more would be sampled separately. The number of perpetrators involved in the case is represented by the Expected number
11
Maximum number of samples, Number of long strips and Number of short strips node
The last event that influences the decision is the maximum number of samples. This Maximum
number of samples node represents the maximal number of samples that could be taken if
everything was sampled as separately as possible. Looking back at Fig. 1 this means that for long strips of tape, where a differentiation into the two original ends and the rest of the back side is possible, the maximum number of samples that can be taken is three (two for each original end and one for the rest of the back side). For a short strip of tape, the maximum number of samples is one, as a short strip denotes a strip that is so small that a differentiation into original ends and the rest of the backside is not possible anymore. As a result, a short strip is sampled whole. Thus, the number of these different sized strips of tape influences the maximum number of samples that can be taken in the investigation. However, it should be noted that in reality assigning the strips of tape to these two categories is not always clear. Often the tape was cut after the incident, and fingermarks trace examiners were not able to see the connections anymore and arrived at an uneven number of original ends. Some adjustments would have to be made in the network to account for these issues. At this point, the Number of short strips node has the states ‘0’, ‘1’, ‘2’, ‘3’ and ‘>3’ and the Number of long
strips node has the states ‘0’, ‘1’, ‘2’, ‘3’, ‘4’, ‘5’ and ‘>5’. These represent the most commonly
encountered numbers in cases. The Maximum number of samples node itself has the states ‘0’ to ‘18’ and ‘>18’ in order to account for all the possible options based on the number of short and long strips of tape.
The maximum number of samples influences the sampling strategy because as soon as that number gets too high1, it would be considered to sample more together in order to reduce the sample size. However, as mentioned before, the effect of the maximum number of samples will be outweighed by other events. The goal is always to conduct the sampling as properly for the case at hand as possible. For example, if masking is expected and a vast maximum number of samples would have to be taken, more locations will still be sampled separately to prevent masking.
The utility and its influences
Utility sampling strategy node
The utility node Utility sampling strategy is depicted in green in Fig. 1 and contains the utility table with all the utilities for each consequence. As utilities are weights based on the personal preference of the decision-maker, DNA trace examiners would have to assign utilities for the established network.
Location of DNA relevant? node
Other than the decision itself, two more parameters influence the utility of the outcomes. The first one is the Location of DNA relevant? node with the states ‘Yes’ and ‘No’. It states whether it was of interest where the DNA was found on the tape. For source level questions, the location of the DNA on the tape is less relevant, as the result will only be used to connect a person to the tape. However, for activity level questions, the location becomes of more interest. Original ends are seen as the location where a perpetrator must have touched the tape while tearing or cutting it. For the utility, this means if these original ends were sampled separately from the rest of the back side and the location was of relevance it would get a higher utility compared to having sampled the entire back side together.
A factor that could further influence the utilities here is the order in which the pieces of tape came from the roll. To determine that order, the physical end matching would be necessary, which could show where the original end was located on the original roll.
1 There is no exact threshold given, but usually a number in the double digits would cause reconsideration of the
Best DNA profiling result node
The second parameter that influences the utilities is the best DNA profiling result that can be obtained through the examination. The Best DNA profiling result node has the states ‘No profile of interest’, ‘Profile that is not suitable for comparison’, ‘Profile that is suitable for manual comparison’ and ‘Profile that is suitable for manual comparison and a database search’. The state ‘No profile of interest’ denotes everything that led to not generating a profile of interest. The second state of having a profile that is not suitable for comparison describes profiles that show some alleles, but which quality was not good enough for manual comparison with reference profiles. The last two states describe profiles that were suitable for manual comparison, and in one state additionally for a database search. The decision on the quality of these profiles is made by the reporting scientist interpreting the results, and categorizing profiles might vary between experts. It is considered that having a profile that is suitable for a database search is of the highest quality. It has to be noted that this node is a summary of the entire item. Irrespective of how often a particular profile was found on the entire tape of a case, this node only states the highest obtained quality. Additionally, it also only concerns the profiles of a person of interest. Profiles of any quality matching with a non-target contributor fall under the state of obtaining no profile of interest.
One striking finding regarding the Best DNA profiling result node is that there are no other nodes connected to it. One might expect that the type of case, the sampling strategy, or the actual number of samples taken in an examination influence the best DNA profiling result that can be obtained. However, conducting Chi-squared tests with the data collected through the case file study showed that these three parameters do not have a significant influence on the best DNA profiling result. One reason for this could be that indeed, there is no correlation between these parameters and the best profiling result. However, the data the case file study was based on stems from real examinations where the trace examiners already sample with their idea of an optimal strategy to get the best result. Therefore, the data may be biased and unable to show these influences. Further investigation would be necessary to examine these influences. The way the network is modeled now, the Best DNA profiling result node includes general success rates stating how often which quality of a profile was obtained.
13
Discussion
During this project, a Bayesian Decision Network for the decision of which sampling strategy to use for sampling contact traces was established. After instantiating the necessary nodes based on the case circumstances, the network can calculate the expected utility for each decision. Therefore, the network provides the expert with the decision that has the highest expected utility. As its calculations are based on actual case file data, it provides a more rational way of choosing the sampling strategy instead of basing the decision on the expert’s opinion and experience[11]. This way, it can reduce bias in the decision-making of trace examiners. As such, it can be used as a guideline for each trace examination or to generally investigate if the currently made decisions coincide with the results of the network. Additionally, the network allows examining the differences between source level and activity level questions. Its node Location of DNA relevant? can be instantiated for both states and the results for the highest expected utility under both states compared. The network can, therefore, highlight the differences in how the tape would need to be sampled in order to gain activity level information while still being most successful.
Adhesive tape and its assessment on activity level
However, during the project, several potential issues with the activity level assessment of adhesive tapes were identified. One of these issues is the step of taking the tape off the object and apart from itself. This step could cause the relocation of DNA within different locations on the tape and would, therefore, make finding DNA at a specific location less reliable. How significant these effects are is not known, and experiments would have to be conducted to investigate this issue.
Moreover, a differentiated sampling of the outer and inner layers of tape that was wrapped around an object could yield valuable information for the assessment on activity level. However, this was not included in the sampling strategies as these layers are currently rarely taken into consideration for sampling by DNA trace examiners. This is due to their challenging and insufficient documentation. Nonetheless, as the sticky side of another layer was on top of the back side that would be sampled the question arises how much DNA can even still be found on that back side. It is fair to assume that the sticky side collects much of the DNA present on the back side and it has to be investigated how much DNA can still be found in these inner layers. Unfortunately, as of yet there is also no non-destructive method to sample the sticky side of the tape for DNA. These issues would have to be investigated to ensure that the different sampling strategies applied by the trace examiners yield reliable information for the assessment on activity level.
Evaluation of the interviews showed that within one discipline, there are no significant differences between the trace examiners’ answers. The only difference was found in how the DNA trace examiners would sample for an object case. Accordingly, the case file study showed that there is no uniform sampling strategy for object cases. Any of the mentioned sampling strategies were applied. The main factor leading to this discrepancy is that it is not clear whether to sample for activity level or source level and how successful it would be to sample more locations separately. The BDN constructed in this study can assist in making this decision in the future. However, the main issue lies in the requests made by the police and prosecution. The initial requests are usually made on source level, whereas the evaluation in the courtroom will always take place on activity and offense level. Therefore, trace examiners presently keep activity level in mind and try to get as much information for the activity level assessment out of the examination. However, this might not always be necessary based on the reason the police and prosecution sent in the piece of evidence. For a proper examination, it should be clear from the beginning what the aim of the examination is. The presented network can aid in making the complex examination of adhesive tapes more transparent. It can show to the other
members of the criminal justice system what is already possible during the examination of the tape and how important it is to think about activity level beforehand. Furthermore, it can show what case information is actually needed in order to conduct a proper trace examination, and therefore, also reduce possible concerns about bias through context information[11].
Adjustments in the established network
Regarding the aim and relevance of the police and prosecution requesting an examination of the tape, it would be possible to add nodes about their decision-making in the network. Their assigned relevance of the item to the case could influence the trace examiners’ examination. In general, a useful feature of the Bayesian Decision Network is that it can be easily adjusted and expanded if necessary. In the entire examination protocol of adhesive tapes, a multitude of decisions have to be made, and some of these decisions might influence the network established in this project. Once those decisions are analyzed using Bayesian decision theory, the separate networks could be combined to show a network of the entire protocol.
Furthermore, discussions with reporting scientists for DNA evidence and experts from other disciplines revealed additional influences that might be necessary to fully assess the utility of the decision of the sampling strategy. For example, some reporting scientists would base the utilities for obtaining a profile suitable for a database search or a profile only suitable for manual comparison on whether a suspect is known from the beginning or not. If a suspect is already known, there might be no difference in preference between these two qualities. In that instance, one wants to link the suspect to the tape, which can be done with either of these qualities. However, if no suspect is known, a profile suitable for a database search would get a higher utility than a profile that is only suitable for manual comparison, as this gives the chance to find a suspect through the database search. On the other hand, some argue that having a profile suitable for a database search is always better as it might result in a higher likelihood ratio for the match due to its better quality.
Another example would be whether the number of samples actually taken in the examination has an influence on assigning utilities. Trace examiners, as the decision-makers for our decision at hand, argue that they do not consider it as they are more concerned about conducting a proper examination than reducing the sample size. Reporting scientists and other members of the criminal justice system, however, might be more concerned about reducing the sample size and, therefore, time and resources. It would have to be established if the network should indeed only show the decision-makers utilities or if all the other mechanisms that are involved should also be considered. This only shows how complex such a decision problem is and how important it is to analyze it. It would be impossible to expect trace examiners to be aware of all these different factors for every piece of evidence. The analysis of such a decision problem using Bayesian decision theory and additionally modeling a BDN could help trace examiners making better-informed decisions.
Conclusion
In conclusion, the network established in this project is a first step in showing how complex decisions can be simplified and made more explicit using Bayesian decision theory and networks. In theory, such a BDN could be modeled for every type of evidence that trace examiners deal with and even for every step in the analysis of evidence (from the crime scene investigation to the evaluation). Nevertheless, as also shown, analyzing a decision with Bayesian decision theory is a challenging task due to its complexity. More issues within the protocol and communication may arise during the analysis. However, these are also valuable insights that can only assist in optimizing the entire workflow in the criminal justice system.
15
Acknowledgments
I want to thank the Netherlands Forensic Institute for giving me the chance to conduct this research at their facility and allowing me to look into their workflow and protocols. Also, a big thank you to all the trace examiners of human biological traces and fingermarks, the reporting scientists on human biological traces, and experts from other departments who gave me the opportunity to talk to them and get insight into their work.
Special thanks to Yvonne van de Wal, Bas Kokshoorn, Marjan Sjerps, and Marcelle Vos for their impeccable supervision during this entire project.
References
[1] I. Helsloot, J. Groenendaal, Naturalistic decision making in forensic science: Toward a better understanding of decision making by forensic team leaders, J. Forensic Sci. (2011). https://doi.org/10.1111/j.1556-4029.2011.01714.x.
[2] A. Biedermann, S. Bozza, F. Taroni, Analysing and exemplifying forensic conclusion criteria in terms of Bayesian decision theory, Sci. Justice. (2018). https://doi.org/10.1016/j.scijus.2017.07.002.
[3] A. Biedermann, S. Bozza, F. Taroni, Decision theoretic properties of forensic identification: Underlying logic and argumentative implications, Forensic Sci. Int. (2008). https://doi.org/10.1016/j.forsciint.2007.11.008.
[4] F. Taroni, S. Bozza, C. Aitken, Decision Analysis in Forensic Science, J. Forensic Sci. (2005). https://doi.org/10.1520/jfs2004443.
[5] R. Wieten, J. De Zoete, B. Blankers, B. Kokshoorn, The interpretation of traces found on adhesive tapes, Law, Probab. Risk. (2015). https://doi.org/10.1093/lpr/mgv012.
[6] S.M. Bleay, R.S. Croxton, M. De Puit, Fingerprint Development Techniques: Theory and Application, 2018. https://doi.org/10.1002/9781119187400.
[7] R. Cook, I.W. Evett, G. Jackson, P.J. Jones, J.A. Lambert, A hierarchy of propositions: Deciding which level to address in casework, Sci. Justice - J. Forensic Sci. Soc. (1998). https://doi.org/10.1016/S1355-0306(98)72117-3.
[8] A. Biedermann, C. Champod, G. Jackson, P. Gill, D. Taylor, J. Butler, N. Morling, T. Hicks, J. Vuille, F. Taroni, Evaluation of forensic DNA traces when propositions of interest relate to activities: Analysis and discussion of recurrent concerns, Front. Genet. (2016). https://doi.org/10.3389/fgene.2016.00215.
[9] HUGIN EXPERT. HUGIN Researcher, HUGIN Manual GUI, (2017). http://download.hugin.com/webdocs/manuals/Htmlhelp/ (accessed July 22, 2019).
[10] S.R. Johnson, G.A. Tomlinson, G.A. Hawker, J.T. Granton, B.M. Feldman, Methods to elicit beliefs for Bayesian priors: a systematic review, J. Clin. Epidemiol. (2010). https://doi.org/10.1016/j.jclinepi.2009.06.003.
[11] I.E. Dror, Human expert performance in forensic decision making: Seven different sources of bias†, Aust. J. Forensic Sci. (2017). https://doi.org/10.1080/00450618.2017.1281348. [12] B. Bhoelai, B.J. de Jong, M. de Puit, T. Sijen, Effect of common fingerprint detection techniques on subsequent STR profiling, Forensic Sci. Int. Genet. Suppl. Ser. (2011). https://doi.org/10.1016/j.fsigss.2011.09.076.
[13] J.L. Ravanat, T. Douki, J. Cadet, Direct and indirect effects of UV radiation on DNA and its components, J. Photochem. Photobiol. B Biol. (2001). https://doi.org/10.1016/S1011-1344(01)00206-8.
17
Supplementary materials
Appendix A: Flowchart of the examination protocol of adhesive tapes
General information
Based on the information obtained through reading the existing protocol and past examinations, talking to experts from different disciplines involved in the examination of tape, and interviewing trace examiners for fingermarks and human biological traces, a flowchart of the current process was generated (see below). This flowchart depicts the different steps that are taken during the examination of adhesive tapes. It focuses on the tasks of fingermarks trace examiners (shown in blue) and DNA trace examiners (shown in orange). Any other trace (shown in purple) is not illustrated in further detail. However, a document with more information was devised for internal use at the NFI. It contains a summary for the execution of each step, possible interferences this step has with any other trace that can be recovered from tape, and a description of all the decisions involved and their influences. The text highlighted in yellow in the flowchart portrays the different decisions that have to be made at each step. Dotted lines within the flowchart represent conditional steps. These steps will only be conducted if a particular circumstance occurred (e.g., photos of fingermarks are only taken if a fingermark was found).
Recommendations
A few interesting points and recommendations could be drawn from working on the protocol. It is known that cyanoacrylate fuming allows for successful DNA profiling[12]; however, this only applies when the time between the treatment with cyanoacrylate and subsequent DNA sampling is relatively short. Trace examiners report that after one day, the cyanoacrylate forms a hard layer on top of the entire back side of the tape. It is not known if dry cyanoacrylate interferes with the success of DNA sampling, and experiments to investigate its influence are recommended. This concern becomes important when fingermarks were found which have to be photographed. Based on the workload of fingermark photographers, this step sometimes still takes place on the same day or can take up to a month. In general, the entire step of fingermark photography raises many questions. The handling of the tape varies every time, and it is not documented what is precisely done. Photographers might use UV light if they think it is necessary to take a proper picture. They are aware of the effect of UV light on DNA[13] and try to reduce the length of use. However, depending on the number of fingermarks and the surface of the tape, it can take longer than 30 minutes to take a picture. This length of use can already cause degradation of DNA. Better documentation and training in DNA transfer and contamination issues could assist in making the decisions.
The use of liquid nitrogen in order to remove the tape from plastic bags or itself could potentially cause further interference with DNA. Liquid nitrogen is rarely used but will be used in cases where no access to underlying layers of tape could be gained. Through case file studies, it was found that for the five items that were treated with liquid nitrogen, no profile suitable for comparison could be obtained. It should be investigated whether the use of liquid nitrogen caused this or not. If it is caused by liquid nitrogen, it could be reasoned not to conduct DNA sampling after the use of liquid nitrogen anymore, and therefore reduce invested resources and time.
Fingermarks on the back side and sticky side, as well as fragments on the sticky side, are sampled individually for DNA. If a vast number of these traces were found on the tape, the question arises if they should all be sampled or if some of them could be sampled together in order to reduce the sample size. A recommendation here would be to wait with the DNA sampling for the results of the database search for fingermarks. This way, fingermarks that match victims and witnesses could be excluded. Additionally, if a suspect is found through the fingermarks database search, it could even be decided to omit the step of sampling the fingermarks for DNA. This would reduce the number of samples taken and therefore reduce resources and time. However, better communication between the police who conduct the fingermarks database search and the Netherlands Forensic Institute would be needed. For more information on the overall success of sampling fingermarks and fragments for DNA, see
Appendix C.
Generally, reporting scientists expressed interest in conducting the examination in stages: conduct a first round of sampling and wait for its results; if the results did not provide the necessary information for the case, another round of sampling could be conducted. This can be applied to any trace that can be found on tape and could, therefore, reduce time and resources. However, if another round of sampling has to be conducted, the piece of evidence has to be retaken out of its packaging. Every handling step can cause loss of DNA or transfer of DNA from one location to another. So the usefulness of sampling in stages for DNA would need to be assessed. Other options to reduce invested time and resources include sampling everything in the first round of the examination but not analyzing it yet or to pool DNA extracts of several locations if they by themselves did not yield results.
The last recommendation concerns the step of taking the tape off the object and apart from itself. In order for DNA trace examiners to conduct a differentiated sampling of the outer layer and inner layer of the tape, better documentation would be needed. Currently, it is difficult for fingermarks trace examiners to document these layers and simultaneously pay attention to not excessively destroying and touching the tape. DNA trace examiners expressed an interest in joining this step in order to see for themselves what might be of importance to sample. This way, DNA trace examiners could also assist in determining what the original ends are. Nevertheless, determining the original ends poses a difficult task depending on the type of tape and whether the perpetrator cut the tape or not. A suggestion that arose is that experts from the physical end matching would join and help to determine the original ends. This could help to ensure that only relevant locations are sampled for DNA.
19
Trace examination of adhesive tapes
Request & assessment of case
Photo packaging, taking out of packaging & overview photos evidence
Other traces (blood, bitemarks, hair, fibers, explosive residue)
White light & alternative light sources (entire item)
1st round cyanoacrylate fuming & checking with white light (back side)
FM found: photo
UV?
Taking tape off object/pulling apart FM found: photo
Sampling overlapping FM for DNA
All traces again after taking tape off/apart; blood additionally after 2nd round cyanoacrylate fuming possible.
Quantity?
Cutting out piece for chemical composition analysis
2nd round cyanoacrylate fuming & checking with white light (back side)
Sampling DNA contact traces
Gentian violet staining (sticky side)
Examination done FM found: photo
Sampling FM for DNA
UV? Location?
Small strip tape: yes or no?
Original ends/back side? Length original end? Separately/together?
Marking fragments together?
Sampling before taking tape off/apart possible!
FM found: areas without FM can be sampled before photos!
FM found: photo
Sampling FM & fragments for DNA Quantity?
Original end: partly/entirely?
Cut? Liquid nitrogen? Stop? Original ends? Key: = DNA = Fingermarks = Other disciplines = Conditional step to = Decision FM = Fingermark
Appendix B: Results and recommendations from interviewing trace examiners
General conclusions from the interviews
Most of the information obtained in this study is based on the interviews conducted with the DNA and fingermarks trace examiners. While this is necessary as they are the decision-makers, bias is a factor that always has to be considered. There are several different sources for bias[10], of which most are unavoidable. Each trace examiner has his/her own values and experience and will base their given answers on that. Furthermore, the interviews were conducted in a closed environment, which means that they could talk to each other in between the interviews. This can influence their thoughts on a certain topic and therefore influence the given answers. The effect of bias on the decisions was not further investigated in this project. Additionally, the interviews were hypothetical situations. Much information was lacking to which the trace examiners would usually have access in a real case. For example, the fingermarks trace examiners would communicate their results of finding fragments on the back side, which could influence the sampling of DNA trace examiners. In the interviews, that information was missing as it is not documented where fragments where found. Therefore, this could have influenced the accuracy of the given answers.
One point that stood out is that most people think in their disciplines and want to conduct the investigation in a way that is best for their trace. While this is understandable, it might not always be the best way to proceed in a case. In order to optimize this a better communication and understanding between the different disciplines but also with the police and prosecution is needed. Police and prosecution would need to state from the beginning what precisely they want to achieve with the investigation, especially concerning source and activity level. That way, the different disciplines that conduct the examination could find the best strategy to answer that question together. Currently, the different disciplines write their own separate reports on the same piece of evidence. If the same question has to be answered with the same disciplines, it could be useful to write a collective report and combine the evidence to answer the question. This way, it would not be up to the judge anymore to combine the results. Recommendations for activity level
As already mentioned there are several issues in the protocol that would influence the evaluation on activity level: possible relocation of DNA while taking the tape apart, the insufficient documentation and knowledge on trace dynamics of layers and the rarity of conducting physical end matching to determine the order the pieces of tape came off the role. Nevertheless, another round of interviews was conducted with DNA trace examiners to gain more insight into how the tape could be sampled for activity level questions. The most common scenarios that are brought forward by the defense were collected from past evaluations on activity level. These scenarios include: the suspect has nothing to do with the alleged activity and never came into contact with the tape, there was innocent contact before the alleged activity (e.g., owner of the tape), secondary transfer, and there was innocent contact with the tape after the alleged activity (e.g., helped to untie a victim). The prosecution scenario always stays the same, stating that the suspect committed the alleged activity.
The results of these interviews revealed that as expected for activity level questions sampling more locations separately becomes more important. Original ends are targeted as it is expected that the person who used the tape must have touched the original ends while tearing or cutting it. However, for the scenario that the suspect never came into contact with the tape, finding the suspect on the tape alone is already enough. The different locations become less
21 end matching could be helpful in order to determine how far inside the roll the DNA was found.
Trace examiners themselves would only sample more locations separately. Scenarios regarding secondary transfer are difficult to verify by sampling, as the DNA of the suspect would be expected at the same locations as the person’s DNA that actually committed the activity. Trace examiners would use the same strategy as if no secondary transfer scenario is given. Lastly, for scenarios regarding innocent contact after the alleged activity targeted sampling of areas that were said to be touched (e.g., areas that were cut) or differentiated sampling of the outer and inner layer can be done in order to verify the statements.
These interviews were based on information that is generally provided years after the examination was already conducted. Therefore, a strategy would have to be used that can account for all of these scenarios from the beginning. This is, however, very challenging. The only option is to sample as many locations, including the outer and inner layer, as separate as possible. After sampling several locations separately, the physical end matching could yield additional information. This shows how important it is that the police and prosecution are already aware of how differently the tape would have to be sampled for source and activity level questions and that they communicate their expectations and goal of the examination. However, this also contains the issue that a suspect is rarely known at the time of the examination. If no suspect is known yet, the investigation is still in its investigative phase, which means that the tape should be examined with the aim of yielding a suspect through a database search. In such a case, sampling more locations together could be more successful. However, a suspect can also be found through tactical information, after which the aim of the examination would shift again. It is, therefore, crucial that the police inform about these developments to ensure an optimal examination of the tape.
Lastly, to increase the amount of information useful for an evaluation on activity level, different sampling techniques for original ends were proposed. The different possibilities can be seen in Fig. 3. The first image shows the currently applied technique, which is an area of about 2 to 3 cm from the original end. The second technique suggests sampling more at the top and bottom rim of the tape, as it should mimic the positions where someone would touch the tape while ripping it. This way, more surface area is sampled, which could provide a higher amount of DNA that is sampled. The last technique leaves out the area close to the upper and lower rim of the tape. This technique is suggested for scenarios where innocent contact with the tape roll is claimed. Logically speaking, all of these techniques could be useful for different instances. Their actual usefulness for different activity level scenarios would need to be tested.
Figure 3: Different sampling techniques for original ends. Depicted in grey is the tape. Black areas show the sampled area for each technique.
Appendix C: Results of the case file study
Success rates on contact traces
In this section, additional success rates from the case file study will be presented. Data on 516 samples from 147 items (investigated tapes) in 80 cases were collected. For evaluating the DNA profiling results, 20 samples and 12 items were excluded as they were either not reported on or not analyzed. This resulted in 496 samples from 135 items for the evaluation.
The general success rates for all 496 samples for each quality of a profile can be seen in Fig. 4. 3% of all samples yielded a profile that was suitable for manual comparison and a database search, and 4% yielded a profile only suitable for manual comparison. 20% resulted in a profile that was not suitable for comparison, and 17% generated a profile that was suitable for manual comparison but only matched with non-target contributors. 56% of all samples generated no profile. Overall, 7% of samples resulted in suitable profiles of interest, while 93% resulted in no suitable profile of interest. Of the 14 profiles suitable for a database search, three resulted in a hit in the database.
Fig. 5 shows the results for the best quality profile. The data was collected for all 135 items and depicts a summary for each item. Irrespective of how often a particular profile was found on the entire item, these results only state the highest obtained quality. Additionally, it also only concerns profiles of a person of interest; profiles of any quality matching with a non-target contributor fall under the category of having no profile of interest. Looking at the results, 10% yielded a profile suitable for manual comparison and a database search, 7% a profile only suitable for manual comparison, 33% a profile that was not suitable for comparison, and 50% no profile. Overall, 17% of items resulted in suitable profiles of interest, while 83% resulted in no suitable profile of interest.
Figure 4: Quality of profile of interest. The pie chart shows the distribution of the different qualities of profiles obtained over all samples. Profiles suitable for a database search are depicted in dark green, profiles suitable for manual comparison in light green, profiles that were not suitable for comparison in yellow, profiles of any quality that only matched with non-target contributors in light grey and no profiles in dark grey.
3% 4%
20%
17% 56%
Quality of profile of interest
DB MC
Not suitable profile Only non-target No profile
23
Figure 5: Best quality profile. The pie chart shows the distribution of the different best qualities of profiles obtained over all items. Profiles suitable for a database search are depicted in dark green, profiles suitable for manual comparison in light green, profiles that were not suitable for comparison in yellow and no profiles of interest in grey. Only profiles of people of interest are considered.
The results of DNA profiling for each sample were analyzed per type of case (see Fig. 6). Of the 496 samples, 40 stem from post-explosion cases, 270 from object cases, and 186 from victim cases. In post-explosion cases, 20% of the samples yielded suitable profiles of a person of interest, in object cases 4% and in victim cases 8%. Additionally, 40% of all samples in victim cases only yielded profiles matching with the victim. Applying Chi-squared tests, these differences were found to be significant. Object cases have a lower chance of generating a suitable profile than explosion and victim cases. Regarding the high success of post-explosion cases, it should be noted that the data also included seemingly intact pieces of tape. This is because the trace examiners rely on the information given by the police. In some instances the police sent in several burnt pieces of tape with seemingly unaffected pieces of tape but reported them as being found in the same location. It is, therefore, possible that pieces of tape were sampled that were not directly subjected to the explosion.
10% 7%
33% 50%
Best quality profile
DB MC
Not suitable profile No profile of interest
Figure 6: Profiles per type of case. The bar diagram shows the distribution of the different qualities of profiles obtained per sample in the three types of cases. ‘E’ represents post-explosion cases, ‘O’ object cases, and ‘V’ victim cases. Profiles suitable for a database search are depicted in dark green, profiles suitable for manual comparison in light green, profiles that were not suitable for comparison in yellow, profiles of any quality that only matched with non-target contributors in light grey and no profiles in dark grey.
The results of DNA profiling for each sample were additionally evaluated per sampling location (see Fig. 7). Of the 496 samples, 29 were taken from the rest of the back side, 128 over the entire back side of the tape, 24 from locations on the back side opposite fingermarks or fragments found on the sticky side (only taken if back side is not already sampled before) and 315 from original ends. For the rest of the back side, 3,5% of the samples provided a suitable profile of a person of interest, sampling the entire back side 12,5%, locations opposite fingermarks and fragments on the sticky side 0% and original ends 5%. Sampling the entire back side together is, therefore, most successful, but results in a loss of activity level information.
Analyzing these results per type of case showed that for the entire back side of the tape, an equal amount of suitable profiles of interest is found in object and post-explosion cases. It represents the only location sampled for post-explosion cases. In object cases, each of the four possible locations was sampled, and sampling the entire back side together provides almost all suitable profiles of interest and is, therefore, most successful. The rest of the backside was only sampled in object cases and resulted once in a suitable profile of interest. In victim cases no suitable profiles were obtained from sampling the entire back side together. The location on the back side opposite found fingermarks and fragments on the sticky side never yielded a profile in object cases (no such samples were taken in post-explosion cases). In victim cases most profiles from that location matched with the victim or did not yield a profile. In one sample a non-usable profile was obtained. This leads to the conclusion that this location has an extremely low chance of providing a suitable profile of interest and could be omitted in the protocol. Regarding the original ends almost all suitable profiles of interest were obtained in victim cases. In victim cases it was also the only location that yielded suitable profiles of interest, and it can, therefore, be concluded that sampling original ends in victim cases is most successful. 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% E O V
Profiles per type of case
No profile Only non-target Not suitable profile MC
25
Figure 7: Profiles per sampling location. The bar diagram shows the distribution of the different qualities of profiles obtained per sample based on the different sampling locations. ‘B’ represents the rest of the back side, ‘E’ the entire back side of the tape, ‘FM’ the back side opposite of a fingermark or fragment that was found on the sticky side and ‘OE’ the original ends. Profiles suitable for a database search are depicted in dark green, profiles suitable for manual comparison in light green, profiles that were not suitable for comparison in yellow, profiles of any quality that only matched with non-target contributors in light grey and no profiles in dark grey.
A last interesting result was discovered in a case where no sampling of the back side of the tape occurred. Instead, the entire tape was cut into pieces and used for DNA extraction. Therefore, these samples contained DNA from the back side as well as the sticky side of the tape. The 16 strips of tape in the case were analyzed individually. Out of these 16 samples, 11 yielded suitable profiles of interest, of which one was suitable for a database search. This shows how successful DNA profiling of the sticky side of the tape can be and supports the importance of finding a non-destructive method for sampling the sticky side of adhesive tapes. Success rates on fingermarks
Additional to the contact traces, the success of finding and sampling fingermarks and fragments was analyzed.
From the 147 items that were evaluated, fingermarks on the back side were found on 8% of the items, fingermarks on the sticky side on 7%, and fragments on the back side and sticky side on 29% and 16% of items respectively (see Fig. 8). This shows that fragments are found more often on items than fingermarks. Especially fragments on the back side are found most often but are currently not considered for sampling. The type of case has no significant influence on how often fingermark traces are found on items.
Analyzing the presence of fragments on the back side showed a significant influence on the best DNA profiling result. If fragments were found, 30% of items yielded suitable profiles of interest, whereas if no fragments were found, only 12% yielded suitable profiles of interest. However, the exact location of these fragments is not documented. Therefore, no conclusion on the influence between a specific location that was sampled and the possible presence of fragments at that location can be drawn.
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% B E FM OE
Profile per sampling location
No profile Only non-target Not suitable profile MC
Figure 8: Found fingermark traces per item. The bar diagram shows the percentages of items on which fingermarks and fragments were found and not found on the back side and sticky side of the tape. The percentage of items where no traces were found is shown in red and the percentage of items where traces were found in green.
Of these numbers, a total amount of 25 fingermarks were found on the back side of the tape, 41 fingermarks on the sticky side, and 56 fragments on the sticky side. Fragments on the back side are not documented. Of the 25 fingermarks on the back side, 32% are located at an original end, while 68% are located somewhere in the middle of the strip of tape. For the fingermarks on the sticky side, 22% are located at an original end and 78% in the middle. In contrast, 62% of fragments on the sticky side are located at an original end, while 38% are located in the middle. Most fingermarks can be found somewhere in the middle of the tape, while fragments occur most commonly on the original ends.
Lastly, the different fingermark traces were evaluated based on their DNA profiling results. Fig. 9 shows that between 10% to 16% of fingermarks and fragments yielded suitable profiles of interest. It can also be seen that fingermarks and fragments on the sticky side generate more profiles than fingermarks on the back side. This would support the hypothesis that more DNA can be found on the sticky side of the tape.
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Fingermarks back side Fragments back side Fingermarks sticky side Fragments sticky side
Found fingermark traces per item
none found found
