Monday, September 9, 2013

Pointing Gestures and False Alerts: Recent Research Suggests How Dogs May Turn into Walking Search Warrants

Publication Note:  A law review article on  Florida v. Harris, written by L.E. Papet and me, has been posted on the website of the Michigan State University Animal Legal and Historical Center. The article makes use of the same phrase, walking search warrants, which we first published here.  The article was the lead in the 2014 issue of the Journal of Animal and Natural Resource Law, a refereed journal of Michigan State University.  Walking Search Warrants: Canine Forensics and Police Culture after Florida v. Harris.

Eighteen police dog teams entered a church where there were no drugs, and where residual odors were unlikely, yet 17 of the 18 teams alerted to the presence of target odors, most multiple times.  In a paper that caused a good deal of consternation in the police dog world, three scientists at the University of California at Davis argued that the false alerts arose from the fact that the police dog handlers had been led to believe that there were drugs inside the church.  The handlers had cued their dogs.  The results were not videotaped and the authors of the paper did not state exactly how the dogs had been induced to alert when their training should have prevented this.  Recent research on human pointing gestures and dog responses goes some way towards explaining how this may have happened. 

In order that drug dogs not become walking search warrants—as an Ohio deputy sheriff described his dog in a recent newspaper article—it is important that supervisors, handlers, and lawyers understand this research and consider its implications. 

Detailed Sweeps  

Handlers of drug dogs take them into areas where drugs may be present and often want to focus the attention of a dog on a particular place where drugs may be hidden by using a “detailed sweep.”  The dogs are trained to alert to the odor of specific drugs and that behavior involves the dog exhibiting a trained final response, such as sitting and staring, which the dog was initially trained to do in order to receive a reward if a target odor is in fact present.  A problem arises if the dog anticipates the reward or interprets the handler’s behavior, either by subtle suggestion or overt command, and demonstrates the trained final response regardless of the presence of any target odor.  (For a general discussion of cueing, see Cueing and Probable Cause, a periodically updated electronic article on the website of the Animal Legal and Historical Center of the Michigan State University College of Law.)

If the handler uses pointing gestures to direct the dog’s attention to a particular location, which is routinely done in detailed sweeps, the question then becomes whether the dog uses that information solely as a direction to smell in an area, or if the gesture itself is being interpreted in part as a command to alert.  In the latter event, the dog has been cued, and the alert should not provide probable cause for any subsequent investigation by the police who are present.  A recent study published in Animal Cognition tells us something about this complex issue, and it is worth the attention of both police dog handlers and lawyers involved in drug cases where dogs were part of the process by which the drugs were found. 

Dogs React to Pointing Gestures

A person hides a reward in one of two identical cups.  A dog enters the room where the food has been hidden and the person points at one of the cups.  Dogs, more often than any other species that has been tested, will go towards the cup which the person is pointing at, more than chimpanzees, our closest living relative, and more than wolves, of which dogs are a subspecies. By six weeks of age, puppies can follow a human pointing gesture even if this means moving away from the human’s hand.  Four scientists at the Max-Planck Institute for Evolutionary Biology in Leipzig, Germany, describe this skill as special and evidence that “selection pressures during domestication may have affected dogs’ ability to use human communication.” 

But do dogs see a human pointing gesture as “an informative communicative act” or as a command?  The four scientists, Linda Scheider, Juliane Kaminski, Josep Call, and Michael Tomasello, in a paper published in the scientific journal, Animal Cognition, designed an experiment to try to answer this question.  They did not reach a definitive conclusion, but their findings advance the debate and are worth careful consideration.   

Earlier Study Where Dogs Ignored Their Noses

In 2003, another team of scientists from the Department of Ethology of Eötvös Lóránd University in Budapest, had found that when confronted with a choice of two bowls, dogs could ignore their own better knowledge and follow the pointing gesture of a human.  One bowl contained strong-smelling food, potent enough for the dog to be able to smell it from its starting position.  The other bowl was empty. In 79% of cases dogs followed the pointing gesture, suggesting to that team that they were interpreting the pointing as a command. The Hungarian team concluded:

“In the case of contradictory cues, dogs prefer to rely on the human communicative signaling (pointing) when they have only olfactory information about the hiding place. Their willingness to do this decreases, however, if they themselves are in the position to obtain visual information about the ‘state of the world’. This means that if a dog sees where the food is placed, it is more reluctant to go in the opposite direction pointed at by the human, even if they had some ever-day experience (with their owner for example) that pointing strongly correlates with the presence of food. This suggests that dogs do not follow human pointing blindly; they seem to have some control over their response to the pointing gesture. However, it is interesting to note that in both studies about half of dogs in these experimental groups seem to fall in either the two following categories. Some dogs mostly ‘believed’ their own eyes, whilst others would go to the empty bowl indicated by the pointing. This also suggests that, possibly due to social experience for some dogs, human pointing becomes one of the most reliable sources of information in the environment.”

This suggests that in choosing dogs to train for detection work, an effort should be made to find dogs that “believe their own eyes.”  This earlier research, it is worth noting, included Ádam Miklósi, perhaps the most important animal behaviorist currently working on canine cognition.    

Experiment 1

The team from the Max-Planck Institute devised two experiments.  In the first, 96, 50 females and 46 males, of various breeds were used.  Dogs were pets and ranged from 1 to 13 years.  To ensure that dogs could make only one choice, even when a human was not present, cups were placed in a long Plexiglas box with an opening to a compartment at each end.  There was a distance of 1.2 meters between openings.  A sliding cover had to be moved to reach the cup inside the compartment, and once moved the other opening was closed, confining the dog to one choice per trial. 

The experimenter, a 27-year-old woman, caught a dog’s attention by clicking her tongue, then baited one of the cups in full view of the dog so that the dog would get used to the idea that there would be food during subsequent trials.  If the dog chose the correct cup, it was allowed to eat the food.  If it chose the wrong cup, it got no reward.  Food was hidden on both sides so that the dog would know that there were two possible locations for food. 

Dogs were divided into groups where the experimenter made the pointing gesture but added no ostensive cues, and those for whom the experimenter would say things like “Luna, pass’ mal auf; pass’ auf, Luna!” =, in English: “Luna, pay attention; pay attention, Luna!” The procedure that followed this action depended on four conditions: 
  1. Authority leaves: dog knowledgeable. After baiting a cup but pointing to the empty cup, the experimenter left the room and the helper let the dog go. The dog had one minute to choose.
  2. Authority leaves: dog ignorant. The experimenter showed the dog a piece of food but the helper closed a curtain so that the dog could not see the experimenter putting the food in one of the cups.  The experimenter pretended to bait the second cup to eliminate the chance that the dog would receive audible information.  The helper opened the curtain and the experimenter pointed.  Again the dog had a minute to choose.
  3. Authority stays: dog knowledgeable.  This was the same as the first trial, but the experimenter remained in the room after pointing, standing motionless with arms hanging down, head bowed, eyes open.
  4. Authority stays: dog ignorant.  This was the same as the second trial except the experimenter remained. 
Dogs followed the pointing gesture to the empty cup significantly more if they did not know the real location of the food.  When they had seen the cup being baited, they generally relied on their visual experience and chose the baited cup, ignoring the pointing gesture.  Whether or not the experimenter—the pointer—remained present when the choice was made had no effect on the dogs.  Nevertheless, if the experimenter pointed, they made the wrong choice significantly more often (about 30% of the time) than in the control condition, where the experimenter did not point.  The use of the additional verbal cues, such as the words of encouragement, did not significantly alter the results.  The fact dogs followed the pointing gesture 30% of the time, when they should have known where the food was, indicated to the researchers that dogs may indeed “interpret pointing to some extent as a command.” This, it should be noted, is a reason for using more than a few blank trials in training regimens, as a dog's taking of pointing as a command to alert will not be apparent without a sufficient amount of testing. 

Experiment 2

In the second experiment of the Max-Planck team, the authority of the human pointing was varied.  “Authority,” according to the team’s definition, was “a person who is able to control the behavior of dogs in a directive way.”  Thus, sometimes the pointer was an adult, sometimes a child.  If pointing is a command, adults should be obeyed more often.  If pointing is informative, it should not matter who is doing it. 

Here, the person pointing could be an adult female or a boy or girl between 4½ and 5½ years old.  Some of the pointing was “honest,” meaning that the experimenter pointed to the cup with food, and sometimes “deceptive,” with the experimenter pointing to the cup without food.  In honest trials, the dog did not witness the cup being bated.  In deceptive trials, dogs witnessed the baiting and knew where the food was. 

Forty-six dogs participated in these trials.  The researchers recruited 26 mother-child pairs as experimenters, none of which had major contact with the dogs before.  Of the children, 15 were girls and 11 boys.  For safety reasons, both children and adults were separated from the dogs by a Plexiglas wall.  Two opaque plastic cups were placed on a wooden board with a distance of 1.3 meters between them.  Most mother-child pairs tested two dogs. 

The researchers found that the dogs “did not differentiate between children’s and adults’ pointing gestures.  The dogs followed the pointing gesture and found the food irrespective of the authority level of the person pointing.”  Nevertheless, when given explicit commands by adults and children to sit, the dogs largely ignored commands by children.  This suggests, according to the researchers “that dogs do not interpret pointing as a strong command comparable to a command like, e.g., ‘sit’.”  This means that anyone analyzing whether a narcotics detection dog is being cued by a handler’s pointing gestures should not expect results as automatic as when a dog sits, stands, or lies down on command.  This also argues that, in establishing testing regimens (whether during training or for certification), there should be a significant number of blank trials in which no target odor is present, as a small number of trials may not reveal a tendency on the part of a dog to alert as a result of pointing gestures.  

Significance of the Research for Cueing Issues in Criminal Trials

The Max-Planck research team concludes that “pointing is a gesture that dogs mainly choose to ignore in situations in which they have better knowledge.”  Not all their results indicated that dogs ignore their own senses when a human’s pointing gesture would lead them to an empty cup.  Thus, their results were not as stark in finding that pointing overcomes knowledge as were the 2003 results of Szetei et al., which the current researchers say might be due to the fact that the earlier research involved both olfactory and visual modalities, while their research “exclusively addressed the visual modality. Seeing food and then following another visual stimulus (the gesture) to an alternative location may be more difficult than smelling the food and then following a cue based in another modality, i.e., visual.”  This observation is important for cueing arguments regarding narcotics detection dogs in that it suggests that pointing may overcome a dog’s reliance on its sense of smell even more than its sense of sight.  Ideally, a narcotics detection dog should not allow a pointing gesture to induce a trained response that is supposed to be solely to a smell—the dog’s alert, yet with a dog that has begun to rely too much on a handler’s pointing gestures, this may be exactly what is happening. 

These behavioral studies used pet dogs that had not been subjected to rigorous scent specific training.  They had not been required to undergo repetitive activities involving pointing, particularly frequent in the early stages of drug dog training.  Such instructional pointing is used in part as a command for the dog to smell in a certain area, but it is important that it not become a command to exhibit the behavior that indicates the presence of drugs.  (For a series of still photographs showing how a pointing gesture can be easily turned into a command, see Ensminger and Papet (2011).  How to Prevent Cueing Arguments from Getting Canine Evidence Thrown Out in Court.  Deputy and Court Officer, 3(2), 36-39.  Unfortunately, it does not appear that the photographs associated with our article are available online so a library copy will have to be obtained, or it may be available from personnel in some courthouses.) 

If a drug dog goes where directed, but does not alert without smelling the odor of drugs, then there is no harm. The question then becomes, for an officer trying to assure the reliability of his dog, or a defense counsel trying to attack the alert of a drug dog, whether the dog tends to alert more often when there is a pointing gesture than when there is not, and to have false positives that cannot be so easily explained away by residual odors.  This requires that there be accurate training records, and that those records involve a significant number of blank trials. (Field records may also be helpful, such as when compiling overall statistics and in comparing teams over significant periods of time.) 

Cueing and the U.S. Supreme Court

The concept of cueing is mentioned in the Supreme Court’s decision in Harris, where Justice Kagan states that “even assuming a dog is generally reliable, circumstances surrounding a particular alert may undermine the case for probable cause—if, say the officer cued the dog (consciously or not), or if the team was working under unfamiliar conditions.”  Among things that defense counsel should look for are indications that pointing has, for a dog, become more than informative, that it has become a command that the dog alert. 

Justice Kagan said that a defendant “must have an opportunity to challenge such evidence of a dog’s reliability [from certification or training programs], whether by cross-examining the testifying officer or by introducing his own fact or expert witnesses.” Professor Andrew Taslitz recently argued—correctly in our opinion—that the Court’s reference to expert testimony can “be read to include expert scientific testimony, and the risk of cuing is indeed one about which the science of dog detection warns.” Such studies as the ones described here are therefore important for counsel to consider in developing defense arguments.  (Professor Taslitz’s recent article appears in the American Bar Association’s publication, Criminal Justice, and may be downloaded from the magazine’s website.)


Fortunately for those who wish to determine a possible propensity for alerting because of a handler’s conscious or unconscious signals, more is now being learned about canine responses to human behavior patterns.  We will be adding a number of important behavioral studies to our running article on Cueing and Probable Cause that are significant in this connection.  We are also writing an analysis of Harris and Jardines where we will discuss the impact the Supreme Court’s decisions are likely to have on police practices and culture. Meanwhile, research such as that discussed here provides important guidance for supervisors concerned with handlers whose dogs’ alerts are too often leading to pointless investigations, and defense counsel who want to question whether a dog was reliable enough to base probable cause on its alert.   

  1. Lit, L., Schweitzer, J.B., and Oberbauer, A.M. (2011).  Handler Beliefs Affect Scent Detection Dog Outcomes. Animal Cognition, 14(3), 387-394.
  2. Scheider, L., Kaminski, J., Call, J., and Tomasello, M. (2013). Do Domestic Dogs Interpret Pointing as a Command?  Animal Cognition, 16(3), 361-372.
  3. Szetei, V., Miklósi, A., Topal, J., and Csanyi, V. (2003).  When Dogs Seem to Lose Their Nose: An Investigation on the Use of Visual and Olfactory Cues in Communicative Context Between Dog and Owner.  Applied Animal Behaviour Science, 83(2), 141-152.
  4. Taslitz, A.E. (Summer 2013). The Cold Nose Might Actually Know?  Science and Scent Lineups.  Criminal Justice, 28(2), 4-8, 55-7.
 This blog was written by John Ensminger and L.E. Papet. 

No comments:

Post a Comment