Thesis, March 1997, by Adee Schoon, Rijksuniversiteit Leiden




In a scent identification line-up a trained dog matches the scent a perpetrator, left on a crime-related object (the corpus delicti), to the scent of a suspect. Scent identifications are performed in a number of countries but the experimental protocol varies and the results are used in different ways. In the Netherlands, scent identifications are used by the police during their investigation, and are also used as evidence in court. In the current experimental protocol, the odour of the suspect is one of twelve different odours, presented in two rows of six. The dog is given the scent of the perpetrator on the corpus delicti, and has to compare it with the odours in the rows. The row containing the odour of the suspect may be the first or the second row the dog is sent to, and the position of the suspect's odour in that row is random. The handler does not know which odour belongs to the suspect. When the dog only responds to the odour of the suspect and ignores all the other odours, this is a "positive identification". The results of a scent identification line-up are recorded in an official report. The reliability of these line-ups has been the subject of debate for several years. This led to a study-project funded by the Ministry of Justice, with this thesis as the result.


Reliability was defined as follows:

1) How often is a suspect correctly identified as the perpetrator in cases where the suspect was, in fact, the perpetrator;

2) How often is a suspect incorrectly identified as the perpetrator in cases where the suspect was not, in fact, the perpetrator?

Since one does not know the "reality" in case studies, an empirical approach was chosen and the study was based on experiments that simulated forensic reality. Practically all the experiments were conducted with operational police tracker dogs. This imposed some limits on the experiments.


In the first part of the thesis some necessary background information is given. The history of scent identification procedures in The Netherlands is outlined in chapter 1. Changes in the course of this century are described in the light of problems encountered: both training-problems and discussions in court led to changes in scent identification procedures. The historical perspective leads to a better understanding of the current scent identification line-up. Human scent identifications by dogs are based upon a number of assumptions: that each human being has a unique odour that dogs can recognise, and that this odour remains stable in time. Two mini-review chapters are therefore included, one describing the current knowledge of odour perception (chapter 2), and one on what is known of human odour production and its perception by dogs (chapter 3). The basic assumptions underlying scent line-ups are supported by the studies reviewed.


Part 2 of the thesis describes the status quo at the beginning of the study. The way the dogs had been trained to perform scent identifications is described in chapter 4. The basic question during the first two years was to find factors that had an effect on the reliability of the identifications, and for the first experiments the corpus delicti was varied in a number of ways (chapter 5). The general results were poor. Analyses of the results showed that neither the type of material of the corpus delicti, nor the amount of time it was handled by the perpetrator, seemed to have a significant effect on the results (within the limits of the experiments). Dogs did not seem to use olfactory information on sex or smoking habits in narrowing down their choice, but did sometimes have a preference for a particular position in the row.


Dogs were able to match odours from different body sites of the same person to each other, and it was clear that the results of matches that were practised were better (chapter 6). Forensic practice and training were critically reviewed in chapter 7.

In an evaluation of the results (chapter 8), a number of reasons for non-recognition's and for incorrect responses were proposed. The conclusion was that although dogs are capable of performing scent identification line-ups, the current experimental protocol could not meet the problems encountered during this study. The focus of the study changed from examining the effect of variation in the corpus delicti to examining the effect of changes in method.


Part 3 contains the results of variations of experimental design, conducted in the latter 2,5 years of the project. Diagnostic ratio's, i.e. the ratio between the percentage of correct identifications in suspect = perpetrator cases and the percentage incorrect identifications in suspect not= perpetrator cases, are used as a method for measuring reliability (chapter 9). The results of the variations in method are described in the last four chapters.


In chapter 10, four experimental set-ups were compared using eight operational police tracker dogs that were trained to retrieve stainless steel tubes containing the odour of the suspect, selecting it from a row of such tubes each containing a different odour. The experiments showed that the "positive check" method, which combined a motivation/ ability check of the dog and a check for a possible preference the dog may have for the odour of the suspect, led to significantly better results than the current method.


In a second study (chapter 11) the same four experimental set-ups were compared, but now with four dogs that were trained to respond to odours on pieces of cloth contained in glass jars. The results showed the same trend as in the former study, and also indicated that the level of false responses was lower in the cloth-responding method than in the tube-retrieving method. This was reasoned to be the result of the difference between the response of the dog (barking, lying down) and its reward (retrieving game) in the cloth-responding method, whereas in the tube-retrieving method this difference is less marked as retrieving a tube is both response and reward.


An improved "positive check" method was chosen as experimental set-up for a first assessment of reliability, which is described in chapter 12. The method of such an assessment is described: it is necessary to perform experiments simulating situations where the suspect is the same as the perpetrator who handled the corpus delicti, but also experiments where the suspect is not the perpetrator. The results of such an assessment with six operational dogs were evaluated, and compared with the results of proficiency testing of other forensic investigation tools.


In chapter 13, a completely new developed method of scent identifications is described. All the scent line-ups described up to now in this thesis basically follow a "match-to-sample" procedure, where the dogs have to match the odour of the perpetrator on the corpus delicti (the sample) to that of the suspect by selecting it out of a number of alternatives (the match). However, in forensic cases, this cannot always be done: the suspect may be completely innocent, or may not have handled the corpus delicti. By employing an "odd-even" procedure, this problem is circumvented: two odours are the same i.e. the suspect is the perpetrator, or two odours differ i.e. the suspect is not the perpetrator. In a pilot study a training method and an experimental design were developed and four dogs were trained in this protocol. The results of this pilot study indicate that the method is a viable one.


In conclusion, the achievement of this thesis has been to create a framework for further research. A number of different ideas to improve the performance of the dogs have been subjected to preliminary research. Not all questions have been answered and further improvement in the reliability through new methods and by an enhanced quality-guarding scheme seem possible. The original questions concerning the reliability of scent identification had a legal background, and in conclusion some implications of the results for the different players in the legal system are enumerated. Awareness of scientific protocol and a continuous quality-guarding scheme are key elements for the police force. Courts should be aware of differences in reliability between different forensic research methods.


For further information or a copy of the Thesis contact:

Dr. Adee Schoon
Rijksuniversiteit Leiden
Institute for Evolutionary and Ecological Sciences
Ethology section
PO box 9516
2300 RA Leiden
The Netherlands

Fax: + (31) 71 5274900