SCENT IDENTIFICATIONS BY DOGS (CANIS FAMILIARIS): A NEW EXPERIMENTAL DESIGN

Traditionally, match-to-sample designs are used for forensic scent identifications: the scent of a perpetrator on a corpus delicti is matched to the scent of a suspect. In a number of cases, e.g. when the suspect is innocent, no match is possible, which leads to specific difficulties. In a new design an odd-even paradigm was followed, and care was taken to meet forensic prerequisites in the experimental setup. Four dogs were trained to compare a human scent (odour 1) on stainless steel tubes, training objects, or typical forensic objects to a human scent on a stainless steel tube (odour 2). Comparisons could be either 'odd' (1 ≠ 2) or 'even' (1 = 2). If the dogs performed poorly in the beginning of an experimental series, they were disqualified from making forensically interesting comparisons. Realistic experiments demonstrated the ability of dogs to compare scents following this protocol, but also showed that the results were influenced by the type of odour 1 and by the type of comparison. The performance of the dogs is compared to the performance of operational dogs in a match-to-sample design: the level of matching 'even' scents is comparable, but the level of non-matching in 'odd' comparisons is substantially higher in the new design. Scent identifications following an odd-even paradigm seem to be more reliable than the customary design. Introducing this new design would however require significant changes in attitude and working conditions of the police.