The Mythic Infallibility Of The

Humans have appreciated the olfactory abilities of dogs since antiquity and have developed various breeds to capitalize on these abilities. The use of dogs to assist in hunt ing is the oldest and probably most familiar example, but trained dogs are also used in various law enforcement tasks such as tracking fugitives and sniffing out narcotics. For example, James Earl Ray, who was convicted of killing Martin Luther King, Jr., escaped from the Brushy Mountain State Penitentiary in Tennessee but was tracked by two bloodhounds and quickly returned to prison. Dogs are also used by the police to identify suspects in lineups. Just as human witnesses to a crime may be asked to select a matching suspect from a lineup, a trained dog may be given a tool or piece of clothing from a crime scene to smell and then be presented with a lineup that includes a suspect and several other people. If the dog shows by it s behav ior that the odor of the suspect matches that of the object from the crime scene, testimony to this effect by the dog's handler is typically admissible in a trial. In fact, Andrew Taslitz reported in 1990 that thousands of these lineups have been done in the United States since 1920, contributing to convictions for robbery, rape, and murder and sentences up to life imprisonment or death.

What is the legal basis for admitting evidence from scent lineups in trials? According to Taslitz, this basis is remarkably weak, at least in the United States. Most court s use four criteria, which were developed for deciding the validity of tracking and simply transplanted to scent lineups: that the dogs should belong to a breed "characterized by acuteness of scent and the power to discriminate among individual human beings" (1990:120), that they be trained for tracking, that there be evidence of their reliability in tracking, and that there be other evidence independent of the dogs that is consistent with the scent identification. In many cases, convictions have been based primarily on scent identification by trained dogs, with additional evidence being only circumstantial. In practice, courts have relied almost exclusively on the claims of handlers about the abilities of their dogs. For example, in one robbery case, a handler testified that a tracking dog named Bobby was 100% accurate in training, as well as in four previous criminal cases. Both the original court and an appeals court accepted this testimony as "ample proof of reliability to justify admitting the results of Bobby's tracking" (1990:55).

Taslitz suggests that the faith of judges and juries in scent identification by dogs is rooted in a kind of mystical belief in man's best friend. Our culture has numerous legends, as well as true stories, attesting to the loyalty of dogs. In Homer's Odyssey, for example, Odysseus returns home after 20 years and is recognized by his aged dog Argus, who dies in the process of greeting his master. Argus probably recognized Odysseus after this long absence at least partly by smell. Because dogs are known for their loyalty, honesty, and integrity, we tend to accept evidence based on scent identification by dogs relatively uncritically. By contrast, imagine using cats, with their reputation for deviousness, in this way.

Because of "the mythic infallibility of the dog," in Taslitz's (1990:20) words, defense lawyers have so far been unsuccessful in persuading courts to apply the Frye Rule to evidence from tracking or scent lineups. The Frye Rule states that a scient ific principle should be well established and generally accepted by scient ists in the appropriate field for expert testimony based on that principle to be used as evidence in court. The purpose of the Frye Rule is to prevent juries from being swayed by the testimony of expert s when the scientific foundat ion for that testimony is weak or nonexistent. In several cases, defense attorneys have argued that conv ictions should be overt urned because the Fr ye Rule was not applied to the testimony of dog handlers. Judges have given various reasons for denying this argument, including the claim that such identification is not based on science, so the Fr ye Rule doesn't apply. This is clearly faulty logic. The olfactory abilities of dogs are subject to experimental testing, and a few such experiments have in fact been done (Brisbin et al. 2000). What is the evidence that trained dogs can recognize unique odors of individual people and use this ability to accurately identify subjects in lineups?1

In an interesting study with twins, Peter Hepper (1988) tested the hypothesis that humans have indiv idual odors that can be recognized by dogs. Hepper was particularly interested in the role of genetic and environmental factors in causing people to have different odors. Therefore, he used three sets of twins in his experiments. The first set consisted of male fraternal twins that were 2 to 3 months old. These twins were genetically different but had a common env ironment because they were being raised in the same home. In particular, they ate the same foods, so any effect of diet on body odor should have been the same for both members of a pair. The second set consisted of male identical twins2 that were 2 to 3 months old. These twins were not only genetically identical but also shared a common environment. The third set consisted of male identical twins that were between 34 and 50 years old, lived separately, and ate different foods. These twins were genetically identical but had different environments.

Four dogs were thoroughly trained in scent discrimination before being used in the experiments. This doesn't mean that the dogs learned how to distinguish between two similar scents; if they had an ability to discriminate, it was probably innate. Instead it means that the dogs were trained to show by their behavior that they were mak ing a choice. The basic protocol was to wash each of the subjects with the same soap, then carefully rinse off the soap (presumably the adult subject s washed themselves). Four T-shirts for each pair of twins were also washed with the same detergent. Each twin then wore one T-shirt for 24 hours and a second T-shirt for the next 24 hours. Finally, a dog was presented with a T-shirt scented by one person, then given a choice between the second T-shirt worn by that person and one of the T-shirts worn by his twin. The dog's handler did not know which was the correct match.

The results of this experiment suggest that dogs can use either genetic or environmental factors to discriminate between people. The average percentage of correct choices was 89% for the infant fraternal twins sharing a common environment, 49% for the infant ident ical t wins sharing a common env iron-ment, and 84% for the adult identical twins living in different environments. Since the dogs were choosing between two T-shirts in each trial, they should have been able to get 50% of their choices correct just by chance. The 49% rate of success at distinguishing twins with the same genes and same environment is not different from chance performance, suggest ing that the dogs got no useful cues for discrimination in this situation. This provides a baseline for comparison with the other two types of trials, where performance of the dogs was significantly better than chance. For the trials with adult identical twins, environmental differences such as diet must have been the basis for discrimination by the dogs. For the trials with infant fraternal twins, genetic differences must have been the basis for discrimination because infant identical twins could not be distinguished.3

Are these results sufficient to validate the use of dogs to identify suspects in scent lineups? Not by a long shot, for at least four reasons (not including the fact that 2-month-olds rarely commit crimes). First and least important, Hepper tested only four dogs, so we can't say if their abilities are common or unusual among dogs in general (this is least important because Hepper did show that some dogs have remarkable abilities to discriminate human odors). Second, the success rates were impressive, but the performance of the dogs was not perfect even after intensive training. Dogs made mistakes in identification 11% of the time in trials with infant fraternal twins and 16% of the time in trials with adult identical twins. Error rates of this magnitude that resulted in convictions of innocent people would be unacceptable. Third, dogs used in forensics often have to compare an odor from one body part of a perpetrator (e.g., head odor on a hat left at the crime scene) with an odor from another body part of a suspect in a lineup (the suspect and several control individuals use their hands to apply scent to a test object such as a metal cylinder). In Hepper's experiments, odors all came from the torsos of the twins. Just as individuals may differ in smell, various body parts of the same individual may also differ. Can dogs be trained to ignore this anatomical variation and detect a component of the olfactory signal that's common to all body parts of an individual, if such a common signal exists? Finally, Hepper's experiment differs from real-world forensic pract ice in that a suspect in a lineup may not be the same as the perpetrator of a crime. If this is the case, a dog given an object from a crime scene to sniff should pick no one from the lineup because no odors of these people would match the smell of the object from the crime scene. In Hepper's experiment, by contrast, one of the T-shirts presented to the dog was always a correct choice.

I. Lehr Brisbin and Steven Austad (1991) did a small experiment to see how the ability of dogs to distinguish between the scents of two different individuals was affected by the body parts that supplied the scent s. They used three dogs and modeled their procedure after that used in competitions authorized by the American Kennel Club. In these competitions, dogs are required to select metal and leather dumbbells scented by their handlers' hands when given a choice between these and dumbbells scented by the judge's hand. This differs from forensic pract ice in an important way because we might expect that it would be easier for dogs to distinguish between the familiar odors of their handlers and the unfamiliar odors of judges than between the odors of various people in l ineups, which would all be equally unfamiliar. This means that the experiment of Brisbin and Austad gives the benefit of the doubt to the hypothesis that dogs can generalize across body parts to identify individuals. If their dogs were able to do this, it might be because the handlers' odors were so familiar, which would provide only weak evidence that dogs could generalize in scent lineups. However, if the dogs used by Brisbin and Austad were not able to generalize, it seems unlikely that dogs could do so in the more challenging circumstances of the lineups.

The dogs were trained by using standard guidelines of the American Kennel Club for obedience training of "utility dogs" for at least 6 months, then further trained in the specific procedures used by Brisbin and Austad in their experiment s. In each trial, two scented dumbbells were placed on a board about 10 feet from the dog and handler while the dog and handler were facing in the opposite direction. When the dumbbells were in place, the dog and handler were instructed to turn around to face the board. The handler gave the dog a command to go to the board. Sniffs of each of the dumbbells were recorded, as well as which dumbbell was retrieved. The experimenters positioned the dumbbells in such a way that the handler didn't know which was the correct choice.

The dumbbells were scented by having a person hold a dumbbell in his or her hand for 30 seconds or using tongs to position the dumbbell in the interior crook of the elbow and having the person hold it tightly there for 30 sec onds. Several types of trials were conducted. First, the dogs had to choose between (1) a dumbbell scented by the handler's hand and one with no human scent or (2) a dumbbell scented by the handler's elbow and one with no human scent. The dogs had no trouble with these discrim inations, averaging 93% correct in the first case and 86% correct in the second case. The dogs also had little trouble in the second type of trial, distinguishing a dumbbell scented by the handler's hand from one scented by a stranger's hand, which is comparable to the task used in obedience competitions. Success rates of the three dogs were 69%, 70%, and 90%, for an average of 76%. However, when given a choice bet ween a dumbbell scented by their handler's elbow versus one scented by a stranger's hand, the dogs were less successful. Success rates were 70%, 57%, and 46%, and the average of 58% was not significantly different from the chance performance of 50%. Since the dogs were trained to identify the hand odors of their handlers, they may have been confused by differences between elbow odors of their handlers and hand odors of strangers. It is interesting that they could distinguish elbow and hand odors of their handlers, with an average success rate of 78%.

In a nutshell, these result s suggest that different body parts of the same person have different odors. This shouldn't be surprising, although the fact that the elbow and hand of the same arm smell differently may be a bit curious. More important, the results suggest that dogs do not automatically generalize from one body part of an individual to another to discriminate between two people with individually distinctive odors, if the dogs are trained by using standard methods. It's quite conceivable that a training regime could be devised to improve the performance of dogs in this task, but present methods that are used not only to train dogs for competitions sponsored by the American Kennel Club but also for scent ident ification in police work are not adequate. This undermines one of the fundamental assumptions of the use of evidence from scent lineups in court.

A group of researchers led by Ray Settle of the Police Dog Training School in Preston, England, developed and tested a training rout ine that they thought might be more effective than standard methods of training dogs to generalize across body parts in identifying individuals (Settle et al. 1994). Settle's group used seven dogs of various breeds. They collected body scents from more than 700 volunteers from various schools, a local business, and a nursing home. Each volunteer was given four pieces of cotton cloth that had been washed and placed in a glass jar. The volunteer was asked to place each piece of cloth on a different part of his or her body for 30 minutes, then replace the scented cloths in the jar. The choice of body parts to be scented was up to the volunteers.

The dogs were tested by giving them one cloth from one of the volunteers to sniff and then either presenting them with a group containing another cloth scented by the same volunteer plus five cloths scented by five other volunteers or presenting them with six steel tubes that had been held by volunteers for 5 minutes. These procedures are similar to those used in act ual scent lineups, at least in Europe. Handlers trained their dogs in a series of progressively more difficult discriminations. For example, in an early step in training, the dogs had to identify a scented cloth when it was placed in the training room with five others that had been washed and handled with tongs, so they had no human scent. In successive steps, the dogs had to discriminate a target scent from one other human scent, then two others, and so on. Training lasted for 9 months.

After this regimen, the dogs were correct 80% of the time on average in the first type of test, in which they had to discriminate among six cloths scented by different people, and 85% of the time in the second type of test, in which they had to discriminate among six steel tubes handled for 5 minutes by different people. In each case, since one of the six choices matched the target scent, we would expect the dogs to be correct one-sixth of the time (17%) purely by chance. Accuracy of 80 to 85% is much better than chance performance, implying that the dogs really had learned to distinguish indiv idual human odors in a situation similar to that used in actual police lineups. Furthermore, Settle and his colleagues suggested that the dogs were ident ifying individuals regardless of whether or not the body part used as a source of the target scent was the same as that used in the lineups.

These results of Settle's group are inconclusive, however, because the volunteers who prov ided cloths scented by four different part s of their bodies handled all of the cloths and placed them together in a closed glass container that was ret urned to the experimenters. In handling the cloths, hand odors may have been transferred to them; in keeping them in closed containers for up to 4 days, odors may have been transferred among them. This means that the odors on four cloths prepared by the same volunteer were probably more similar to each other than, for example, the odors on the dumbbells held in the hand and the elbow by a subject in the experiment of Brisbin and Austad (1991). Because of this likely odor contamination in the study by Settle and his colleagues, it may have been easier for the dogs to match a target scent to one of the scents in the lineup than if it had come from a distinctly different part of the body than the part used to create the lineup. Therefore, this study doesn't restore much faith in the validity of evidence from scent lineups, because it doesn't convincingly show that dogs can generalize from one body part to another to distinguish odors from different people, even though the training used by Settle's group was more extensive than that normally given to police dogs.

The most extensive and most recent set of experiments dealing with scent identificat ion by dogs in police lineups was done by Gertrud Schoon for her dissertation at the University of Leiden in the Netherlands.4 She is affil iated with the Department of Criminalistics and Forensic Science and the Ethology Group of the Inst itute of Evolut ionary and Ecological Sciences at the university, indicat ing that she has broad training and diverse interest s. Real progress in science often comes from people who bring a new perspective to a long-standing problem. In this case, Schoon's background in ethology (ani mal behavior) may have helped her to find creative ways of dealing with some of the practical problems of forensic science that we've been discussing.

Schoon (1996, 1997, 1998) examined several aspects of scent identification to find ways to improve the training of dogs and the operation of scent lineups, but I will discuss just one part of her work that addresses one of the most important pitfalls of these lineups. This is the possibility that a dog will falsely accuse a suspect by select ing the suspect's scent from a lineup when this scent does not match the target scent from an object left at a crime scene. Unlike all the experiments described so far that used a match-to-sample design, in which an odor matching the sample was always present in the array with which the dogs were tested, in the complicated real world of police investigations innocent suspect s are sometimes arrested. In this case, the correct choice of a dog would be to select none of the odors in a lineup because none would match that from the crime scene. But there are various factors that might work against dogs making this response. They might select the odor that was closest to the odor on the target object, even if it wasn't identical. In fact, when human witnesses to crimes are shown pictures of several potential subjects, they tend to pick one that look s most similar to the person they saw at a crime scene, even if that person was not actually there (Wells et al. 2000). In scent lineups, the control scents often come from pol ice officers who may be familiar to the dog, so the dog may pick the odor from the lineup that is least familiar, regardless of whether it matches the scent of the target object. The handler may believe the suspect is guilty and therefore reward the dog for mak ing any select ion at all. If the lineup consists of a suspect and several control individuals, none of whom are known to the handler, the handler may pick out the likely suspect by using v isual cues and communicate this identification to the dog unconsciously.5 Even if the lineup consists of objects like metal cylinders that were scented by using standardized methods, the handler, who is likely to be a police officer familiar with the case, may believe the suspect is guilty and therefore unconsciously encourage the dog to choose one of the cylinders when the correct response if the suspect was really innocent would be no choice. To date, Schoon is the only researcher who has rigorously tested the possibility of false identification of suspects based on dogs and scent lineups.

Schoon did her experiment s with six tracking dogs that were trained for police work in the Netherlands. Each dog was used in 10 trials. In each one, a target scent was prepared by hav ing a volunteer police employee treat an object as if it had been found at a crime scene. These objects were screwdrivers, wrenches, pistol buttplates, sweat shirt cuffs, and scent samples from the seat of the volunteer's car. In preparing these target scent s, the volunteers were acting as if they were the perpetrators of a crime.

To prepare odors for scent lineups, Schoon gave volunteers two glass jars with six stainless steel cylinders in each. The volunteers were instructed to handle the cylinders in each jar for 5 minutes. In addit ion, one of the volunteers handled a piece of polyvinyl chloride (PVC ) tubing, providing a "check"

scent that was used as described below to be sure the dogs didn't have an inherent preference for one of the odors in the lineup but were really making a choice based on matching an odor in the lineup to a target odor. For each trial with a dog, one volunteer was designated as a suspect. In half of the trials with each dog, this "suspect" was the same as the "perpetrator" who had scented the objects from a hypothetical crime scene; in the remaining trials, the suspect and perpetrator were different people.

Each trial consisted of four tests of discrimination in which seven scented cylinders were laid out for a dog. One cylinder had the scent of the suspect, one cylinder had the scent of the person who prov ided the check scent, and the other five cylinders had scents of five other volunteers that Schoon called decoys. In the first two trials, the dog was allowed to smell the check scent on the PVC tubing, then shown the lineup of seven cylinders. In this case, the correct choice was to select the cylinder scented by the person who made the check scent. If the dog made the wrong choice in either of these tests, it was disqualified for that trial. For example, the dog might be disinterested in working that day and make no choice, or it m ight select the suspect's scent, indicating a preference for that scent even though it didn't match the check scent. In these cases, Schoon bel ieved it was invalid to test the dog's ability to compare a perpetrator's and suspect's scent. One of her pragmat ic suggestions is that police department s begin using this kind of performance check in act ual lineups.

After these two trials, the check scent was removed from the lineup and the dog was given the scented object made by the person pretending to be the perpetrator. Then it was given two tests with six scented cylinders—one scented by the simulated suspect and five scented by decoys. Recall that in half of the trials with each of the six dogs, the suspect was the same as the perpetrator, whereas in the other half the suspect and perpetrator were different, and the handler did not k now which trial was which. Since there were five types of objects, there were 10 trials per dog, for a total of 60 trials.

The performance of the dogs in this experiment was not particularly impressive. In 30 of the 60 trials, the dogs were disqualified because they made errors in the check tests. In the remaining trials in which the suspect was the same as the perpetrator, the dogs correctly selected the scent of the suspect in four trials, they selected a decoy's scent in five trials, and they made no selection in two trials, for a success rate of 4/11, or 36%. In the trials in which the suspect was not the same as the perpetrator, the dogs correctly made no selection in nine cases, they selected the suspect's scent in one case, and they selected a decoy's scent in nine cases, for a success rate of 9/19, or 47%. It's interesting that the dogs made some choice in a majority of these trials, suggesting that it may have been difficult for them to resist pick ing one item from the lineup even when there was no match to the scent on the object from the hypothetical crime scene. However, the effective error rate was only 1/19, or 5%, because in nine cases the dogs selected a decoy, who would presumably be known to be innocent in a real lineup. These estimates assume that the conditions of this experiment with simulated suspects and perpetrators are representative of actual forensic practice. Also, the estimates are not very precise because they are based on a small number of successful trials with only six dogs.

Getting Started With Dumbbells

Getting Started With Dumbbells

The use of dumbbells gives you a much more comprehensive strengthening effect because the workout engages your stabilizer muscles, in addition to the muscle you may be pin-pointing. Without all of the belts and artificial stabilizers of a machine, you also engage your core muscles, which are your body's natural stabilizers.

Get My Free Ebook

Post a comment